This is an archive of the discontinued LLVM Phabricator instance.

Differential D106116

GlobalISel: Preserve LLT when bitcasting loads and stores
ClosedPublic

Authored by arsenm on Jul 15 2021, 6:32 PM.

Details

Reviewers
paquette
aemerson
foad
aditya_nandakumar
dsanders
bogner
Summary

This also avoids improperly legalizing some truncating vector stores.

Diff Detail

Unit TestsFailed

TimeTest
3,340 msx64 debian > libarcher.critical::critical.c
3,350 msx64 debian > libarcher.critical::lock.c
3,250 msx64 debian > libarcher.parallel::parallel-firstprivate.c
3,430 msx64 debian > libarcher.parallel::parallel-simple2.c
3,360 msx64 debian > libarcher.races::critical-unrelated.c
View Full Test Results (18 Failed)

Event Timeline

arsenm created this revision.Jul 15 2021, 6:32 PM
Herald added subscribers: kerbowa, hiraditya, rovka and 2 others. · View Herald TranscriptJul 15 2021, 6:32 PM
arsenm requested review of this revision.Jul 15 2021, 6:32 PM
Herald added a project: Restricted Project. · View Herald TranscriptJul 15 2021, 6:32 PM
Herald added a subscriber: wdng. · View Herald Transcript
Harbormaster completed remote builds in B114398: Diff 359187.Jul 15 2021, 7:34 PM
paquette accepted this revision.Jul 16 2021, 10:53 AM

LGTM

This revision is now accepted and ready to land.Jul 16 2021, 10:53 AM
arsenm closed this revision.Jul 19 2021, 8:30 AM

9236125ec88f619799879beef5a6d2855c87bced

Revision Contents

PathSize
llvm/
include/
llvm/
CodeGen/
5 lines
lib/
CodeGen/
GlobalISel/
13 lines
test/
CodeGen/
AMDGPU/
GlobalISel/
2 lines
8 lines
138 lines
84 lines
457 lines
140 lines
16 lines
122 lines
168 lines

Diff 359187

llvm/include/llvm/CodeGen/MachineMemOperand.h

Show First 20 Lines • Show All 323 Lines • ▼ Show 20 Linespublic:
/// Change the SourceValue for this MachineMemOperand. This should only be /// Change the SourceValue for this MachineMemOperand. This should only be
/// used when an object is being relocated and all references to it are being /// used when an object is being relocated and all references to it are being
/// updated. /// updated.
void setValue(const Value *NewSV) { PtrInfo.V = NewSV; } void setValue(const Value *NewSV) { PtrInfo.V = NewSV; }
void setValue(const PseudoSourceValue *NewSV) { PtrInfo.V = NewSV; } void setValue(const PseudoSourceValue *NewSV) { PtrInfo.V = NewSV; }
void setOffset(int64_t NewOffset) { PtrInfo.Offset = NewOffset; } void setOffset(int64_t NewOffset) { PtrInfo.Offset = NewOffset; }
/// Reset the tracked memory type.
void setType(LLT NewTy) {
Lint: Pre-merge checks
Inline Actions

clang-format: please reformat the code

-  void setType(LLT NewTy) {
-    MemoryType = NewTy;
-  }
+  void setType(LLT NewTy) { MemoryType = NewTy; }
Lint: Pre-merge checks: clang-format: please reformat the code ``` - void setType(LLT NewTy) { - MemoryType = NewTy…
MemoryType = NewTy;
}
/// Profile - Gather unique data for the object. /// Profile - Gather unique data for the object.
/// ///
void Profile(FoldingSetNodeID &ID) const; void Profile(FoldingSetNodeID &ID) const;
/// Support for operator<<. /// Support for operator<<.
/// @{ /// @{
void print(raw_ostream &OS, ModuleSlotTracker &MST, void print(raw_ostream &OS, ModuleSlotTracker &MST,
SmallVectorImpl<StringRef> &SSNs, const LLVMContext &Context, SmallVectorImpl<StringRef> &SSNs, const LLVMContext &Context,
Show All 25 Lines

llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp

Show First 20 Lines • Show All 2,975 Lines • ▼ Show 20 Lines
} }
LegalizerHelper::LegalizeResult LegalizerHelper::LegalizeResult
LegalizerHelper::bitcast(MachineInstr &MI, unsigned TypeIdx, LLT CastTy) { LegalizerHelper::bitcast(MachineInstr &MI, unsigned TypeIdx, LLT CastTy) {
switch (MI.getOpcode()) { switch (MI.getOpcode()) {
case TargetOpcode::G_LOAD: { case TargetOpcode::G_LOAD: {
if (TypeIdx != 0) if (TypeIdx != 0)
return UnableToLegalize; return UnableToLegalize;
MachineMemOperand &MMO = **MI.memoperands_begin();
// Not sure how to interpret a bitcast of an extending load.
if (MMO.getMemoryType().getSizeInBits() != CastTy.getSizeInBits())
return UnableToLegalize;
Observer.changingInstr(MI); Observer.changingInstr(MI);
bitcastDst(MI, CastTy, 0); bitcastDst(MI, CastTy, 0);
MMO.setType(CastTy);
Observer.changedInstr(MI); Observer.changedInstr(MI);
return Legalized; return Legalized;
} }
case TargetOpcode::G_STORE: { case TargetOpcode::G_STORE: {
if (TypeIdx != 0) if (TypeIdx != 0)
return UnableToLegalize; return UnableToLegalize;
MachineMemOperand &MMO = **MI.memoperands_begin();
// Not sure how to interpret a bitcast of a truncating store.
if (MMO.getMemoryType().getSizeInBits() != CastTy.getSizeInBits())
return UnableToLegalize;
Observer.changingInstr(MI); Observer.changingInstr(MI);
bitcastSrc(MI, CastTy, 0); bitcastSrc(MI, CastTy, 0);
MMO.setType(CastTy);
Observer.changedInstr(MI); Observer.changedInstr(MI);
return Legalized; return Legalized;
} }
case TargetOpcode::G_SELECT: { case TargetOpcode::G_SELECT: {
if (TypeIdx != 0) if (TypeIdx != 0)
return UnableToLegalize; return UnableToLegalize;
if (MRI.getType(MI.getOperand(1).getReg()).isVector()) { if (MRI.getType(MI.getOperand(1).getReg()).isVector()) {
▲ Show 20 Lines • Show All 4,195 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/artifact-combiner-zext.mir

Show First 20 Lines • Show All 249 Lines • ▼ Show 20 Lines
--- ---
name: test_zext_v2s8_to_v2s32_of_sext_v2s8_to_v2s16 name: test_zext_v2s8_to_v2s32_of_sext_v2s8_to_v2s16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CHECK-LABEL: name: test_zext_v2s8_to_v2s32_of_sext_v2s8_to_v2s16 ; CHECK-LABEL: name: test_zext_v2s8_to_v2s32_of_sext_v2s8_to_v2s16
; CHECK: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CHECK: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CHECK: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), addrspace 1) ; CHECK: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CHECK: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CHECK: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CHECK: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CHECK: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CHECK: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CHECK: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
Show All 15 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-extract-vector-elt.mir

Show First 20 Lines • Show All 1,470 Lines • ▼ Show 20 Lines
name: extract_vector_elt_33_v64p3 name: extract_vector_elt_33_v64p3
body: | body: |
bb.0: bb.0:
liveins: $sgpr0_sgpr1 liveins: $sgpr0_sgpr1
; CHECK-LABEL: name: extract_vector_elt_33_v64p3 ; CHECK-LABEL: name: extract_vector_elt_33_v64p3
; CHECK: [[COPY:%[0-9]+]]:_(p1) = COPY $sgpr0_sgpr1 ; CHECK: [[COPY:%[0-9]+]]:_(p1) = COPY $sgpr0_sgpr1
; CHECK: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x p3>), align 4, addrspace 4) ; CHECK: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s32>), align 4, addrspace 4)
; CHECK: [[BITCAST:%[0-9]+]]:_(<16 x p3>) = G_BITCAST [[LOAD]](<16 x s32>) ; CHECK: [[BITCAST:%[0-9]+]]:_(<16 x p3>) = G_BITCAST [[LOAD]](<16 x s32>)
; CHECK: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 64 ; CHECK: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 64
; CHECK: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; CHECK: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; CHECK: [[LOAD1:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<16 x p3>) from unknown-address + 64, align 4, addrspace 4) ; CHECK: [[LOAD1:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<16 x s32>) from unknown-address + 64, align 4, addrspace 4)
; CHECK: [[BITCAST1:%[0-9]+]]:_(<16 x p3>) = G_BITCAST [[LOAD1]](<16 x s32>) ; CHECK: [[BITCAST1:%[0-9]+]]:_(<16 x p3>) = G_BITCAST [[LOAD1]](<16 x s32>)
; CHECK: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 128 ; CHECK: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 128
; CHECK: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64) ; CHECK: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
; CHECK: [[LOAD2:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[PTR_ADD1]](p1) :: (load (<16 x p3>) from unknown-address + 128, align 4, addrspace 4) ; CHECK: [[LOAD2:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[PTR_ADD1]](p1) :: (load (<16 x s32>) from unknown-address + 128, align 4, addrspace 4)
; CHECK: [[BITCAST2:%[0-9]+]]:_(<16 x p3>) = G_BITCAST [[LOAD2]](<16 x s32>) ; CHECK: [[BITCAST2:%[0-9]+]]:_(<16 x p3>) = G_BITCAST [[LOAD2]](<16 x s32>)
; CHECK: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 192 ; CHECK: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 192
; CHECK: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64) ; CHECK: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
; CHECK: [[LOAD3:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[PTR_ADD2]](p1) :: (load (<16 x p3>) from unknown-address + 192, align 4, addrspace 4) ; CHECK: [[LOAD3:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[PTR_ADD2]](p1) :: (load (<16 x s32>) from unknown-address + 192, align 4, addrspace 4)
; CHECK: [[BITCAST3:%[0-9]+]]:_(<16 x p3>) = G_BITCAST [[LOAD3]](<16 x s32>) ; CHECK: [[BITCAST3:%[0-9]+]]:_(<16 x p3>) = G_BITCAST [[LOAD3]](<16 x s32>)
; CHECK: [[FRAME_INDEX:%[0-9]+]]:_(p5) = G_FRAME_INDEX %stack.0 ; CHECK: [[FRAME_INDEX:%[0-9]+]]:_(p5) = G_FRAME_INDEX %stack.0
; CHECK: [[UV:%[0-9]+]]:_(p3), [[UV1:%[0-9]+]]:_(p3), [[UV2:%[0-9]+]]:_(p3), [[UV3:%[0-9]+]]:_(p3), [[UV4:%[0-9]+]]:_(p3), [[UV5:%[0-9]+]]:_(p3), [[UV6:%[0-9]+]]:_(p3), [[UV7:%[0-9]+]]:_(p3), [[UV8:%[0-9]+]]:_(p3), [[UV9:%[0-9]+]]:_(p3), [[UV10:%[0-9]+]]:_(p3), [[UV11:%[0-9]+]]:_(p3), [[UV12:%[0-9]+]]:_(p3), [[UV13:%[0-9]+]]:_(p3), [[UV14:%[0-9]+]]:_(p3), [[UV15:%[0-9]+]]:_(p3) = G_UNMERGE_VALUES [[BITCAST]](<16 x p3>) ; CHECK: [[UV:%[0-9]+]]:_(p3), [[UV1:%[0-9]+]]:_(p3), [[UV2:%[0-9]+]]:_(p3), [[UV3:%[0-9]+]]:_(p3), [[UV4:%[0-9]+]]:_(p3), [[UV5:%[0-9]+]]:_(p3), [[UV6:%[0-9]+]]:_(p3), [[UV7:%[0-9]+]]:_(p3), [[UV8:%[0-9]+]]:_(p3), [[UV9:%[0-9]+]]:_(p3), [[UV10:%[0-9]+]]:_(p3), [[UV11:%[0-9]+]]:_(p3), [[UV12:%[0-9]+]]:_(p3), [[UV13:%[0-9]+]]:_(p3), [[UV14:%[0-9]+]]:_(p3), [[UV15:%[0-9]+]]:_(p3) = G_UNMERGE_VALUES [[BITCAST]](<16 x p3>)
; CHECK: [[UV16:%[0-9]+]]:_(p3), [[UV17:%[0-9]+]]:_(p3), [[UV18:%[0-9]+]]:_(p3), [[UV19:%[0-9]+]]:_(p3), [[UV20:%[0-9]+]]:_(p3), [[UV21:%[0-9]+]]:_(p3), [[UV22:%[0-9]+]]:_(p3), [[UV23:%[0-9]+]]:_(p3), [[UV24:%[0-9]+]]:_(p3), [[UV25:%[0-9]+]]:_(p3), [[UV26:%[0-9]+]]:_(p3), [[UV27:%[0-9]+]]:_(p3), [[UV28:%[0-9]+]]:_(p3), [[UV29:%[0-9]+]]:_(p3), [[UV30:%[0-9]+]]:_(p3), [[UV31:%[0-9]+]]:_(p3) = G_UNMERGE_VALUES [[BITCAST1]](<16 x p3>) ; CHECK: [[UV16:%[0-9]+]]:_(p3), [[UV17:%[0-9]+]]:_(p3), [[UV18:%[0-9]+]]:_(p3), [[UV19:%[0-9]+]]:_(p3), [[UV20:%[0-9]+]]:_(p3), [[UV21:%[0-9]+]]:_(p3), [[UV22:%[0-9]+]]:_(p3), [[UV23:%[0-9]+]]:_(p3), [[UV24:%[0-9]+]]:_(p3), [[UV25:%[0-9]+]]:_(p3), [[UV26:%[0-9]+]]:_(p3), [[UV27:%[0-9]+]]:_(p3), [[UV28:%[0-9]+]]:_(p3), [[UV29:%[0-9]+]]:_(p3), [[UV30:%[0-9]+]]:_(p3), [[UV31:%[0-9]+]]:_(p3) = G_UNMERGE_VALUES [[BITCAST1]](<16 x p3>)
; CHECK: [[UV32:%[0-9]+]]:_(p3), [[UV33:%[0-9]+]]:_(p3), [[UV34:%[0-9]+]]:_(p3), [[UV35:%[0-9]+]]:_(p3), [[UV36:%[0-9]+]]:_(p3), [[UV37:%[0-9]+]]:_(p3), [[UV38:%[0-9]+]]:_(p3), [[UV39:%[0-9]+]]:_(p3), [[UV40:%[0-9]+]]:_(p3), [[UV41:%[0-9]+]]:_(p3), [[UV42:%[0-9]+]]:_(p3), [[UV43:%[0-9]+]]:_(p3), [[UV44:%[0-9]+]]:_(p3), [[UV45:%[0-9]+]]:_(p3), [[UV46:%[0-9]+]]:_(p3), [[UV47:%[0-9]+]]:_(p3) = G_UNMERGE_VALUES [[BITCAST2]](<16 x p3>) ; CHECK: [[UV32:%[0-9]+]]:_(p3), [[UV33:%[0-9]+]]:_(p3), [[UV34:%[0-9]+]]:_(p3), [[UV35:%[0-9]+]]:_(p3), [[UV36:%[0-9]+]]:_(p3), [[UV37:%[0-9]+]]:_(p3), [[UV38:%[0-9]+]]:_(p3), [[UV39:%[0-9]+]]:_(p3), [[UV40:%[0-9]+]]:_(p3), [[UV41:%[0-9]+]]:_(p3), [[UV42:%[0-9]+]]:_(p3), [[UV43:%[0-9]+]]:_(p3), [[UV44:%[0-9]+]]:_(p3), [[UV45:%[0-9]+]]:_(p3), [[UV46:%[0-9]+]]:_(p3), [[UV47:%[0-9]+]]:_(p3) = G_UNMERGE_VALUES [[BITCAST2]](<16 x p3>)
; CHECK: [[UV48:%[0-9]+]]:_(p3), [[UV49:%[0-9]+]]:_(p3), [[UV50:%[0-9]+]]:_(p3), [[UV51:%[0-9]+]]:_(p3), [[UV52:%[0-9]+]]:_(p3), [[UV53:%[0-9]+]]:_(p3), [[UV54:%[0-9]+]]:_(p3), [[UV55:%[0-9]+]]:_(p3), [[UV56:%[0-9]+]]:_(p3), [[UV57:%[0-9]+]]:_(p3), [[UV58:%[0-9]+]]:_(p3), [[UV59:%[0-9]+]]:_(p3), [[UV60:%[0-9]+]]:_(p3), [[UV61:%[0-9]+]]:_(p3), [[UV62:%[0-9]+]]:_(p3), [[UV63:%[0-9]+]]:_(p3) = G_UNMERGE_VALUES [[BITCAST3]](<16 x p3>) ; CHECK: [[UV48:%[0-9]+]]:_(p3), [[UV49:%[0-9]+]]:_(p3), [[UV50:%[0-9]+]]:_(p3), [[UV51:%[0-9]+]]:_(p3), [[UV52:%[0-9]+]]:_(p3), [[UV53:%[0-9]+]]:_(p3), [[UV54:%[0-9]+]]:_(p3), [[UV55:%[0-9]+]]:_(p3), [[UV56:%[0-9]+]]:_(p3), [[UV57:%[0-9]+]]:_(p3), [[UV58:%[0-9]+]]:_(p3), [[UV59:%[0-9]+]]:_(p3), [[UV60:%[0-9]+]]:_(p3), [[UV61:%[0-9]+]]:_(p3), [[UV62:%[0-9]+]]:_(p3), [[UV63:%[0-9]+]]:_(p3) = G_UNMERGE_VALUES [[BITCAST3]](<16 x p3>)
; CHECK: G_STORE [[UV]](p3), [[FRAME_INDEX]](p5) :: (store (p3) into %stack.0, align 256, addrspace 5) ; CHECK: G_STORE [[UV]](p3), [[FRAME_INDEX]](p5) :: (store (p3) into %stack.0, align 256, addrspace 5)
; CHECK: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; CHECK: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
▲ Show 20 Lines • Show All 757 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 1,079 Lines • ▼ Show 20 Lines
--- ---
name: test_load_constant_s96_align16 name: test_load_constant_s96_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_s96_align16 ; CI-LABEL: name: test_load_constant_s96_align16
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 16, addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 4)
; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; VI-LABEL: name: test_load_constant_s96_align16 ; VI-LABEL: name: test_load_constant_s96_align16
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 16, addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 4)
; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-LABEL: name: test_load_constant_s96_align16 ; GFX9-LABEL: name: test_load_constant_s96_align16
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 16, addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 4)
; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p4) = COPY $vgpr0_vgpr1 %0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 16, addrspace 4) %1:_(s96) = G_LOAD %0 :: (load (s96), align 16, addrspace 4)
$vgpr0_vgpr1_vgpr2 = COPY %1 $vgpr0_vgpr1_vgpr2 = COPY %1
... ...
--- ---
name: test_load_constant_s96_align8 name: test_load_constant_s96_align8
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_s96_align8 ; CI-LABEL: name: test_load_constant_s96_align8
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 8, addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 8, addrspace 4)
; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; VI-LABEL: name: test_load_constant_s96_align8 ; VI-LABEL: name: test_load_constant_s96_align8
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 8, addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 8, addrspace 4)
; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-LABEL: name: test_load_constant_s96_align8 ; GFX9-LABEL: name: test_load_constant_s96_align8
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 8, addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 8, addrspace 4)
; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p4) = COPY $vgpr0_vgpr1 %0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 8, addrspace 4) %1:_(s96) = G_LOAD %0 :: (load (s96), align 8, addrspace 4)
$vgpr0_vgpr1_vgpr2 = COPY %1 $vgpr0_vgpr1_vgpr2 = COPY %1
... ...
--- ---
name: test_load_constant_s96_align4 name: test_load_constant_s96_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_s96_align4 ; CI-LABEL: name: test_load_constant_s96_align4
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 4, addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 4)
; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; VI-LABEL: name: test_load_constant_s96_align4 ; VI-LABEL: name: test_load_constant_s96_align4
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 4, addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 4)
; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-LABEL: name: test_load_constant_s96_align4 ; GFX9-LABEL: name: test_load_constant_s96_align4
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 4, addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 4)
; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p4) = COPY $vgpr0_vgpr1 %0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 4, addrspace 4) %1:_(s96) = G_LOAD %0 :: (load (s96), align 4, addrspace 4)
$vgpr0_vgpr1_vgpr2 = COPY %1 $vgpr0_vgpr1_vgpr2 = COPY %1
... ...
--- ---
▲ Show 20 Lines • Show All 487 Lines • ▼ Show 20 Lines
--- ---
name: test_load_constant_s128_align16 name: test_load_constant_s128_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_s128_align16 ; CI-LABEL: name: test_load_constant_s128_align16
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (s128), addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
; CI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; VI-LABEL: name: test_load_constant_s128_align16 ; VI-LABEL: name: test_load_constant_s128_align16
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (s128), addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX9-LABEL: name: test_load_constant_s128_align16 ; GFX9-LABEL: name: test_load_constant_s128_align16
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (s128), addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p4) = COPY $vgpr0_vgpr1 %0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s128) = G_LOAD %0 :: (load (s128), align 16, addrspace 4) %1:_(s128) = G_LOAD %0 :: (load (s128), align 16, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
name: test_load_constant_s128_align4 name: test_load_constant_s128_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_s128_align4 ; CI-LABEL: name: test_load_constant_s128_align4
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (s128), align 4, addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
; CI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; VI-LABEL: name: test_load_constant_s128_align4 ; VI-LABEL: name: test_load_constant_s128_align4
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (s128), align 4, addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX9-LABEL: name: test_load_constant_s128_align4 ; GFX9-LABEL: name: test_load_constant_s128_align4
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (s128), align 4, addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p4) = COPY $vgpr0_vgpr1 %0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s128) = G_LOAD %0 :: (load (s128), align 4, addrspace 4) %1:_(s128) = G_LOAD %0 :: (load (s128), align 4, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
▲ Show 20 Lines • Show All 316 Lines • ▼ Show 20 Lines
--- ---
name: test_load_constant_s256_align32 name: test_load_constant_s256_align32
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_s256_align32 ; CI-LABEL: name: test_load_constant_s256_align32
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (s256), align 16, addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), align 16, addrspace 4)
; CI: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
; VI-LABEL: name: test_load_constant_s256_align32 ; VI-LABEL: name: test_load_constant_s256_align32
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (s256), align 16, addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), align 16, addrspace 4)
; VI: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
; GFX9-LABEL: name: test_load_constant_s256_align32 ; GFX9-LABEL: name: test_load_constant_s256_align32
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (s256), align 16, addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), align 16, addrspace 4)
; GFX9: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
%0:_(p4) = COPY $vgpr0_vgpr1 %0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s256) = G_LOAD %0 :: (load (s256), align 16, addrspace 4) %1:_(s256) = G_LOAD %0 :: (load (s256), align 16, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
... ...
--- ---
▲ Show 20 Lines • Show All 802 Lines • ▼ Show 20 Lines
--- ---
name: test_load_constant_v2s8_align4 name: test_load_constant_v2s8_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_v2s8_align4 ; CI-LABEL: name: test_load_constant_v2s8_align4
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<2 x s8>), align 4, addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; CI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; CI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; CI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; CI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; CI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32) ; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
; CI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; CI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]] ; CI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]]
; CI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32) ; CI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
; CI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; CI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; CI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; CI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; CI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; CI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; CI: $vgpr0 = COPY [[ANYEXT]](s32) ; CI: $vgpr0 = COPY [[ANYEXT]](s32)
; VI-LABEL: name: test_load_constant_v2s8_align4 ; VI-LABEL: name: test_load_constant_v2s8_align4
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<2 x s8>), align 4, addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; VI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; VI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]] ; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
; VI: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 ; VI: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16) ; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; VI: $vgpr0 = COPY [[ANYEXT]](s32)
; GFX9-LABEL: name: test_load_constant_v2s8_align4 ; GFX9-LABEL: name: test_load_constant_v2s8_align4
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<2 x s8>), align 4, addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; GFX9: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
Show All 15 Lines
--- ---
name: test_load_constant_v2s8_align2 name: test_load_constant_v2s8_align2
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_v2s8_align2 ; CI-LABEL: name: test_load_constant_v2s8_align2
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<2 x s8>), addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; CI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; CI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; CI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; CI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; CI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32) ; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
; CI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; CI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]] ; CI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]]
; CI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32) ; CI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
; CI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; CI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; CI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; CI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; CI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; CI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; CI: $vgpr0 = COPY [[ANYEXT]](s32) ; CI: $vgpr0 = COPY [[ANYEXT]](s32)
; VI-LABEL: name: test_load_constant_v2s8_align2 ; VI-LABEL: name: test_load_constant_v2s8_align2
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<2 x s8>), addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; VI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; VI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]] ; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
; VI: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 ; VI: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16) ; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; VI: $vgpr0 = COPY [[ANYEXT]](s32)
; GFX9-LABEL: name: test_load_constant_v2s8_align2 ; GFX9-LABEL: name: test_load_constant_v2s8_align2
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<2 x s8>), addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; GFX9: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
▲ Show 20 Lines • Show All 222 Lines • ▼ Show 20 Lines
--- ---
name: test_load_constant_v4s8_align4 name: test_load_constant_v4s8_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_v4s8_align4 ; CI-LABEL: name: test_load_constant_v4s8_align4
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<4 x s8>), addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; CI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; CI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; CI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; CI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; CI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; CI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; CI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; CI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; CI: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; VI-LABEL: name: test_load_constant_v4s8_align4 ; VI-LABEL: name: test_load_constant_v4s8_align4
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<4 x s8>), addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; VI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; VI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; GFX9-LABEL: name: test_load_constant_v4s8_align4 ; GFX9-LABEL: name: test_load_constant_v4s8_align4
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<4 x s8>), addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
▲ Show 20 Lines • Show All 161 Lines • ▼ Show 20 Lines
--- ---
name: test_load_constant_v8s8_align8 name: test_load_constant_v8s8_align8
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_v8s8_align8 ; CI-LABEL: name: test_load_constant_v8s8_align8
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s8>), addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
; CI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; CI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 10 Lines bb.0:
; CI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; CI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; CI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; CI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; CI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32) ; CI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32)
; CI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>) ; CI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>)
; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>) ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>)
; CI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; CI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>)
; VI-LABEL: name: test_load_constant_v8s8_align8 ; VI-LABEL: name: test_load_constant_v8s8_align8
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s8>), addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 10 Lines bb.0:
; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; VI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; VI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; VI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32) ; VI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32)
; VI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>) ; VI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>)
; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>) ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>)
; VI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; VI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>)
; GFX9-LABEL: name: test_load_constant_v8s8_align8 ; GFX9-LABEL: name: test_load_constant_v8s8_align8
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s8>), addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 25 Lines
--- ---
name: test_load_constant_v16s8_align16 name: test_load_constant_v16s8_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_v16s8_align16 ; CI-LABEL: name: test_load_constant_v16s8_align16
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s8>), addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
; CI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; CI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 28 Lines bb.0:
; CI: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32) ; CI: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32)
; CI: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32) ; CI: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32)
; CI: [[BUILD_VECTOR3:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32) ; CI: [[BUILD_VECTOR3:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32)
; CI: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR3]](<4 x s32>) ; CI: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR3]](<4 x s32>)
; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>) ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>)
; VI-LABEL: name: test_load_constant_v16s8_align16 ; VI-LABEL: name: test_load_constant_v16s8_align16
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s8>), addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 28 Lines bb.0:
; VI: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32) ; VI: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32)
; VI: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32) ; VI: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32)
; VI: [[BUILD_VECTOR3:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32) ; VI: [[BUILD_VECTOR3:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32)
; VI: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR3]](<4 x s32>) ; VI: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR3]](<4 x s32>)
; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>) ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>)
; GFX9-LABEL: name: test_load_constant_v16s8_align16 ; GFX9-LABEL: name: test_load_constant_v16s8_align16
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s8>), addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
▲ Show 20 Lines • Show All 47 Lines • ▼ Show 20 Lines
--- ---
name: test_load_constant_v32s8_align32 name: test_load_constant_v32s8_align32
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_v32s8_align32 ; CI-LABEL: name: test_load_constant_v32s8_align32
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<32 x s8>), addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
; CI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; CI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>)
; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
▲ Show 20 Lines • Show All 64 Lines • ▼ Show 20 Lines bb.0:
; CI: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR22]](s32) ; CI: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR22]](s32)
; CI: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR23]](s32) ; CI: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR23]](s32)
; CI: [[BUILD_VECTOR7:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32) ; CI: [[BUILD_VECTOR7:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32)
; CI: [[TRUNC7:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR7]](<4 x s32>) ; CI: [[TRUNC7:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR7]](<4 x s32>)
; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>), [[TRUNC4]](<4 x s8>), [[TRUNC5]](<4 x s8>), [[TRUNC6]](<4 x s8>), [[TRUNC7]](<4 x s8>) ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>), [[TRUNC4]](<4 x s8>), [[TRUNC5]](<4 x s8>), [[TRUNC6]](<4 x s8>), [[TRUNC7]](<4 x s8>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>)
; VI-LABEL: name: test_load_constant_v32s8_align32 ; VI-LABEL: name: test_load_constant_v32s8_align32
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<32 x s8>), addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
▲ Show 20 Lines • Show All 64 Lines • ▼ Show 20 Lines bb.0:
; VI: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR22]](s32) ; VI: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR22]](s32)
; VI: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR23]](s32) ; VI: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR23]](s32)
; VI: [[BUILD_VECTOR7:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32) ; VI: [[BUILD_VECTOR7:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32)
; VI: [[TRUNC7:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR7]](<4 x s32>) ; VI: [[TRUNC7:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR7]](<4 x s32>)
; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>), [[TRUNC4]](<4 x s8>), [[TRUNC5]](<4 x s8>), [[TRUNC6]](<4 x s8>), [[TRUNC7]](<4 x s8>) ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>), [[TRUNC4]](<4 x s8>), [[TRUNC5]](<4 x s8>), [[TRUNC6]](<4 x s8>), [[TRUNC7]](<4 x s8>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>)
; GFX9-LABEL: name: test_load_constant_v32s8_align32 ; GFX9-LABEL: name: test_load_constant_v32s8_align32
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<32 x s8>), addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>)
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
▲ Show 20 Lines • Show All 939 Lines • ▼ Show 20 Lines
--- ---
name: test_load_constant_v8s16_align8 name: test_load_constant_v8s16_align8
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_v8s16_align8 ; CI-LABEL: name: test_load_constant_v8s16_align8
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s16>), align 8, addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
; CI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
; VI-LABEL: name: test_load_constant_v8s16_align8 ; VI-LABEL: name: test_load_constant_v8s16_align8
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s16>), align 8, addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
; GFX9-LABEL: name: test_load_constant_v8s16_align8 ; GFX9-LABEL: name: test_load_constant_v8s16_align8
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s16>), align 8, addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
; GFX9: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
%0:_(p4) = COPY $vgpr0_vgpr1 %0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<8 x s16>) = G_LOAD %0 :: (load (<8 x s16>), align 8, addrspace 4) %1:_(<8 x s16>) = G_LOAD %0 :: (load (<8 x s16>), align 8, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
▲ Show 20 Lines • Show All 2,429 Lines • ▼ Show 20 Lines
--- ---
name: test_load_constant_v2s128_align32 name: test_load_constant_v2s128_align32
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_v2s128_align32 ; CI-LABEL: name: test_load_constant_v2s128_align32
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s128>), addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
; CI: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
; VI-LABEL: name: test_load_constant_v2s128_align32 ; VI-LABEL: name: test_load_constant_v2s128_align32
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s128>), addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
; VI: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
; GFX9-LABEL: name: test_load_constant_v2s128_align32 ; GFX9-LABEL: name: test_load_constant_v2s128_align32
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s128>), addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
%0:_(p4) = COPY $vgpr0_vgpr1 %0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s128>) = G_LOAD %0 :: (load (<2 x s128>), align 32, addrspace 4) %1:_(<2 x s128>) = G_LOAD %0 :: (load (<2 x s128>), align 32, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
... ...
--- ---
name: test_load_constant_v2p1_align16 name: test_load_constant_v2p1_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_v2p1_align16 ; CI-LABEL: name: test_load_constant_v2p1_align16
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; VI-LABEL: name: test_load_constant_v2p1_align16 ; VI-LABEL: name: test_load_constant_v2p1_align16
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX9-LABEL: name: test_load_constant_v2p1_align16 ; GFX9-LABEL: name: test_load_constant_v2p1_align16
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p4) = COPY $vgpr0_vgpr1 %0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 16, addrspace 4) %1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 16, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
name: test_load_constant_v2p1_align8 name: test_load_constant_v2p1_align8
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_v2p1_align8 ; CI-LABEL: name: test_load_constant_v2p1_align8
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), align 8, addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; VI-LABEL: name: test_load_constant_v2p1_align8 ; VI-LABEL: name: test_load_constant_v2p1_align8
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), align 8, addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX9-LABEL: name: test_load_constant_v2p1_align8 ; GFX9-LABEL: name: test_load_constant_v2p1_align8
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), align 8, addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p4) = COPY $vgpr0_vgpr1 %0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 8, addrspace 4) %1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 8, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
name: test_load_constant_v2p1_align4 name: test_load_constant_v2p1_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_v2p1_align4 ; CI-LABEL: name: test_load_constant_v2p1_align4
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), align 4, addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; VI-LABEL: name: test_load_constant_v2p1_align4 ; VI-LABEL: name: test_load_constant_v2p1_align4
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), align 4, addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX9-LABEL: name: test_load_constant_v2p1_align4 ; GFX9-LABEL: name: test_load_constant_v2p1_align4
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), align 4, addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p4) = COPY $vgpr0_vgpr1 %0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 4, addrspace 4) %1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 4, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
▲ Show 20 Lines • Show All 1,799 Lines • ▼ Show 20 Lines
--- ---
name: test_extload_constant_v2s96_from_24_align4 name: test_extload_constant_v2s96_from_24_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_extload_constant_v2s96_from_24_align4 ; CI-LABEL: name: test_extload_constant_v2s96_from_24_align4
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 4, addrspace 1) ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 1)
; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; CI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; CI: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64) ; CI: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
; CI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) ; CI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
; CI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; CI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; CI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; CI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; CI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; CI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; CI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; CI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; VI-LABEL: name: test_extload_constant_v2s96_from_24_align4 ; VI-LABEL: name: test_extload_constant_v2s96_from_24_align4
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 4, addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; VI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; VI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; VI: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64) ; VI: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
; VI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) ; VI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; VI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; VI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; GFX9-LABEL: name: test_extload_constant_v2s96_from_24_align4 ; GFX9-LABEL: name: test_extload_constant_v2s96_from_24_align4
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 4, addrspace 1) ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 1)
; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; GFX9: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; GFX9: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64) ; GFX9: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) ; GFX9: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
; GFX9: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; GFX9: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; GFX9: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
%0:_(p4) = COPY $vgpr0_vgpr1 %0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 4, addrspace 1) %1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 4, addrspace 1)
%2:_(s96) = G_EXTRACT %1, 0 %2:_(s96) = G_EXTRACT %1, 0
%3:_(s96) = G_EXTRACT %1, 96 %3:_(s96) = G_EXTRACT %1, 96
$vgpr0_vgpr1_vgpr2 = COPY %2 $vgpr0_vgpr1_vgpr2 = COPY %2
$vgpr3_vgpr4_vgpr5 = COPY %3 $vgpr3_vgpr4_vgpr5 = COPY %3
... ...
--- ---
name: test_extload_constant_v2s96_from_24_align16 name: test_extload_constant_v2s96_from_24_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_extload_constant_v2s96_from_24_align16 ; CI-LABEL: name: test_extload_constant_v2s96_from_24_align16
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 16, addrspace 1) ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 1)
; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; CI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; CI: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64) ; CI: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
; CI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) ; CI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
; CI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; CI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; CI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; CI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; CI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; CI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; CI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; CI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; VI-LABEL: name: test_extload_constant_v2s96_from_24_align16 ; VI-LABEL: name: test_extload_constant_v2s96_from_24_align16
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 16, addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; VI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; VI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; VI: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64) ; VI: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
; VI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) ; VI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; VI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; VI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; GFX9-LABEL: name: test_extload_constant_v2s96_from_24_align16 ; GFX9-LABEL: name: test_extload_constant_v2s96_from_24_align16
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 16, addrspace 1) ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 1)
; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; GFX9: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; GFX9: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64) ; GFX9: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) ; GFX9: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
; GFX9: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; GFX9: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; GFX9: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
%0:_(p4) = COPY $vgpr0_vgpr1 %0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 16, addrspace 1) %1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 16, addrspace 1)
%2:_(s96) = G_EXTRACT %1, 0 %2:_(s96) = G_EXTRACT %1, 0
%3:_(s96) = G_EXTRACT %1, 96 %3:_(s96) = G_EXTRACT %1, 96
$vgpr0_vgpr1_vgpr2 = COPY %2 $vgpr0_vgpr1_vgpr2 = COPY %2
$vgpr3_vgpr4_vgpr5 = COPY %3 $vgpr3_vgpr4_vgpr5 = COPY %3
... ...
--- ---
name: test_load_constant_s512_align32 name: test_load_constant_s512_align32
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_s512_align32 ; CI-LABEL: name: test_load_constant_s512_align32
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (s512), align 32, addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
; CI: [[BITCAST:%[0-9]+]]:_(s512) = G_BITCAST [[LOAD]](<16 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(s512) = G_BITCAST [[LOAD]](<16 x s32>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](s512) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](s512)
; VI-LABEL: name: test_load_constant_s512_align32 ; VI-LABEL: name: test_load_constant_s512_align32
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (s512), align 32, addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
; VI: [[BITCAST:%[0-9]+]]:_(s512) = G_BITCAST [[LOAD]](<16 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s512) = G_BITCAST [[LOAD]](<16 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](s512) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](s512)
; GFX9-LABEL: name: test_load_constant_s512_align32 ; GFX9-LABEL: name: test_load_constant_s512_align32
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (s512), align 32, addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
; GFX9: [[BITCAST:%[0-9]+]]:_(s512) = G_BITCAST [[LOAD]](<16 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s512) = G_BITCAST [[LOAD]](<16 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](s512) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](s512)
%0:_(p4) = COPY $vgpr0_vgpr1 %0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s512) = G_LOAD %0 :: (load (s512), align 32, addrspace 4) %1:_(s512) = G_LOAD %0 :: (load (s512), align 32, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY %1
... ...
--- ---
name: test_load_constant_v4s128_align32 name: test_load_constant_v4s128_align32
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_constant_v4s128_align32 ; CI-LABEL: name: test_load_constant_v4s128_align32
; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s128>), align 32, addrspace 4) ; CI: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
; CI: [[BITCAST:%[0-9]+]]:_(<4 x s128>) = G_BITCAST [[LOAD]](<16 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s128>) = G_BITCAST [[LOAD]](<16 x s32>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](<4 x s128>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](<4 x s128>)
; VI-LABEL: name: test_load_constant_v4s128_align32 ; VI-LABEL: name: test_load_constant_v4s128_align32
; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s128>), align 32, addrspace 4) ; VI: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s128>) = G_BITCAST [[LOAD]](<16 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s128>) = G_BITCAST [[LOAD]](<16 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](<4 x s128>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](<4 x s128>)
; GFX9-LABEL: name: test_load_constant_v4s128_align32 ; GFX9-LABEL: name: test_load_constant_v4s128_align32
; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s128>), align 32, addrspace 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s128>) = G_BITCAST [[LOAD]](<16 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s128>) = G_BITCAST [[LOAD]](<16 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](<4 x s128>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](<4 x s128>)
%0:_(p4) = COPY $vgpr0_vgpr1 %0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<4 x s128>) = G_LOAD %0 :: (load (<4 x s128>), align 32, addrspace 4) %1:_(<4 x s128>) = G_LOAD %0 :: (load (<4 x s128>), align 32, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY %1
... ...

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 801 Lines • ▼ Show 20 Lines
--- ---
name: test_load_flat_s96_align16 name: test_load_flat_s96_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_flat_s96_align16 ; CI-LABEL: name: test_load_flat_s96_align16
; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 16) ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; VI-LABEL: name: test_load_flat_s96_align16 ; VI-LABEL: name: test_load_flat_s96_align16
; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 16) ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-LABEL: name: test_load_flat_s96_align16 ; GFX9-LABEL: name: test_load_flat_s96_align16
; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 16) ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p0) = COPY $vgpr0_vgpr1 %0:_(p0) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 16, addrspace 0) %1:_(s96) = G_LOAD %0 :: (load (s96), align 16, addrspace 0)
$vgpr0_vgpr1_vgpr2 = COPY %1 $vgpr0_vgpr1_vgpr2 = COPY %1
... ...
--- ---
name: test_load_flat_s96_align8 name: test_load_flat_s96_align8
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_flat_s96_align8 ; CI-LABEL: name: test_load_flat_s96_align8
; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 8) ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 8)
; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; VI-LABEL: name: test_load_flat_s96_align8 ; VI-LABEL: name: test_load_flat_s96_align8
; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 8) ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 8)
; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-LABEL: name: test_load_flat_s96_align8 ; GFX9-LABEL: name: test_load_flat_s96_align8
; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 8) ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 8)
; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p0) = COPY $vgpr0_vgpr1 %0:_(p0) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 8, addrspace 0) %1:_(s96) = G_LOAD %0 :: (load (s96), align 8, addrspace 0)
$vgpr0_vgpr1_vgpr2 = COPY %1 $vgpr0_vgpr1_vgpr2 = COPY %1
... ...
--- ---
name: test_load_flat_s96_align4 name: test_load_flat_s96_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_flat_s96_align4 ; CI-LABEL: name: test_load_flat_s96_align4
; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 4) ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; VI-LABEL: name: test_load_flat_s96_align4 ; VI-LABEL: name: test_load_flat_s96_align4
; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 4) ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-LABEL: name: test_load_flat_s96_align4 ; GFX9-LABEL: name: test_load_flat_s96_align4
; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p0) = COPY $vgpr0_vgpr1 %0:_(p0) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 4, addrspace 0) %1:_(s96) = G_LOAD %0 :: (load (s96), align 4, addrspace 0)
$vgpr0_vgpr1_vgpr2 = COPY %1 $vgpr0_vgpr1_vgpr2 = COPY %1
... ...
--- ---
▲ Show 20 Lines • Show All 487 Lines • ▼ Show 20 Lines
--- ---
name: test_load_flat_s128_align16 name: test_load_flat_s128_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_flat_s128_align16 ; CI-LABEL: name: test_load_flat_s128_align16
; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (s128)) ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
; CI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; VI-LABEL: name: test_load_flat_s128_align16 ; VI-LABEL: name: test_load_flat_s128_align16
; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (s128)) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX9-LABEL: name: test_load_flat_s128_align16 ; GFX9-LABEL: name: test_load_flat_s128_align16
; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (s128)) ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p0) = COPY $vgpr0_vgpr1 %0:_(p0) = COPY $vgpr0_vgpr1
%1:_(s128) = G_LOAD %0 :: (load (s128), align 16, addrspace 0) %1:_(s128) = G_LOAD %0 :: (load (s128), align 16, addrspace 0)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
name: test_load_flat_s128_align4 name: test_load_flat_s128_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_flat_s128_align4 ; CI-LABEL: name: test_load_flat_s128_align4
; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (s128), align 4) ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
; CI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; VI-LABEL: name: test_load_flat_s128_align4 ; VI-LABEL: name: test_load_flat_s128_align4
; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (s128), align 4) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX9-LABEL: name: test_load_flat_s128_align4 ; GFX9-LABEL: name: test_load_flat_s128_align4
; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (s128), align 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p0) = COPY $vgpr0_vgpr1 %0:_(p0) = COPY $vgpr0_vgpr1
%1:_(s128) = G_LOAD %0 :: (load (s128), align 4, addrspace 0) %1:_(s128) = G_LOAD %0 :: (load (s128), align 4, addrspace 0)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
▲ Show 20 Lines • Show All 1,157 Lines • ▼ Show 20 Lines
--- ---
name: test_load_flat_v2s8_align4 name: test_load_flat_v2s8_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_flat_v2s8_align4 ; CI-LABEL: name: test_load_flat_v2s8_align4
; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<2 x s8>), align 4) ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; CI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; CI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; CI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; CI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; CI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32) ; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
; CI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; CI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]] ; CI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]]
; CI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32) ; CI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
; CI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; CI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; CI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; CI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; CI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; CI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; CI: $vgpr0 = COPY [[ANYEXT]](s32) ; CI: $vgpr0 = COPY [[ANYEXT]](s32)
; VI-LABEL: name: test_load_flat_v2s8_align4 ; VI-LABEL: name: test_load_flat_v2s8_align4
; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<2 x s8>), align 4) ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; VI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; VI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]] ; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
; VI: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 ; VI: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16) ; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; VI: $vgpr0 = COPY [[ANYEXT]](s32)
; GFX9-LABEL: name: test_load_flat_v2s8_align4 ; GFX9-LABEL: name: test_load_flat_v2s8_align4
; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<2 x s8>), align 4) ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; GFX9: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
Show All 15 Lines
--- ---
name: test_load_flat_v2s8_align2 name: test_load_flat_v2s8_align2
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_flat_v2s8_align2 ; CI-LABEL: name: test_load_flat_v2s8_align2
; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<2 x s8>)) ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; CI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; CI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; CI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; CI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; CI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32) ; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
; CI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; CI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]] ; CI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]]
; CI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32) ; CI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
; CI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; CI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; CI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; CI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; CI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; CI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; CI: $vgpr0 = COPY [[ANYEXT]](s32) ; CI: $vgpr0 = COPY [[ANYEXT]](s32)
; VI-LABEL: name: test_load_flat_v2s8_align2 ; VI-LABEL: name: test_load_flat_v2s8_align2
; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<2 x s8>)) ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; VI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; VI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]] ; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
; VI: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 ; VI: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16) ; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; VI: $vgpr0 = COPY [[ANYEXT]](s32)
; GFX9-LABEL: name: test_load_flat_v2s8_align2 ; GFX9-LABEL: name: test_load_flat_v2s8_align2
; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<2 x s8>)) ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; GFX9: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
▲ Show 20 Lines • Show All 222 Lines • ▼ Show 20 Lines
--- ---
name: test_load_flat_v4s8_align4 name: test_load_flat_v4s8_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_flat_v4s8_align4 ; CI-LABEL: name: test_load_flat_v4s8_align4
; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<4 x s8>)) ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; CI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; CI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; CI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; CI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; CI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; CI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; CI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; CI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; CI: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; VI-LABEL: name: test_load_flat_v4s8_align4 ; VI-LABEL: name: test_load_flat_v4s8_align4
; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<4 x s8>)) ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; VI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; VI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; GFX9-LABEL: name: test_load_flat_v4s8_align4 ; GFX9-LABEL: name: test_load_flat_v4s8_align4
; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<4 x s8>)) ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
▲ Show 20 Lines • Show All 161 Lines • ▼ Show 20 Lines
--- ---
name: test_load_flat_v8s8_align8 name: test_load_flat_v8s8_align8
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_flat_v8s8_align8 ; CI-LABEL: name: test_load_flat_v8s8_align8
; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<8 x s8>)) ; CI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
; CI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; CI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 10 Lines bb.0:
; CI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; CI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; CI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; CI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; CI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32) ; CI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32)
; CI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>) ; CI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>)
; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>) ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>)
; CI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; CI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>)
; VI-LABEL: name: test_load_flat_v8s8_align8 ; VI-LABEL: name: test_load_flat_v8s8_align8
; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<8 x s8>)) ; VI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 10 Lines bb.0:
; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; VI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; VI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; VI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32) ; VI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32)
; VI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>) ; VI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>)
; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>) ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>)
; VI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; VI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>)
; GFX9-LABEL: name: test_load_flat_v8s8_align8 ; GFX9-LABEL: name: test_load_flat_v8s8_align8
; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<8 x s8>)) ; GFX9: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 25 Lines
--- ---
name: test_load_flat_v16s8_align16 name: test_load_flat_v16s8_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_flat_v16s8_align16 ; CI-LABEL: name: test_load_flat_v16s8_align16
; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<16 x s8>)) ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
; CI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; CI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 28 Lines bb.0:
; CI: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32) ; CI: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32)
; CI: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32) ; CI: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32)
; CI: [[BUILD_VECTOR3:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32) ; CI: [[BUILD_VECTOR3:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32)
; CI: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR3]](<4 x s32>) ; CI: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR3]](<4 x s32>)
; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>) ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>)
; VI-LABEL: name: test_load_flat_v16s8_align16 ; VI-LABEL: name: test_load_flat_v16s8_align16
; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<16 x s8>)) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 28 Lines bb.0:
; VI: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32) ; VI: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32)
; VI: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32) ; VI: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32)
; VI: [[BUILD_VECTOR3:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32) ; VI: [[BUILD_VECTOR3:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32)
; VI: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR3]](<4 x s32>) ; VI: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR3]](<4 x s32>)
; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>) ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>)
; GFX9-LABEL: name: test_load_flat_v16s8_align16 ; GFX9-LABEL: name: test_load_flat_v16s8_align16
; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<16 x s8>)) ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
▲ Show 20 Lines • Show All 1,177 Lines • ▼ Show 20 Lines
--- ---
name: test_load_flat_v8s16_align8 name: test_load_flat_v8s16_align8
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_flat_v8s16_align8 ; CI-LABEL: name: test_load_flat_v8s16_align8
; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<8 x s16>), align 8) ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
; CI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
; VI-LABEL: name: test_load_flat_v8s16_align8 ; VI-LABEL: name: test_load_flat_v8s16_align8
; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<8 x s16>), align 8) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
; GFX9-LABEL: name: test_load_flat_v8s16_align8 ; GFX9-LABEL: name: test_load_flat_v8s16_align8
; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<8 x s16>), align 8) ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
; GFX9: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
%0:_(p0) = COPY $vgpr0_vgpr1 %0:_(p0) = COPY $vgpr0_vgpr1
%1:_(<8 x s16>) = G_LOAD %0 :: (load (<8 x s16>), align 8, addrspace 0) %1:_(<8 x s16>) = G_LOAD %0 :: (load (<8 x s16>), align 8, addrspace 0)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
▲ Show 20 Lines • Show All 2,281 Lines • ▼ Show 20 Lines
--- ---
name: test_load_flat_v2p1_align16 name: test_load_flat_v2p1_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_flat_v2p1_align16 ; CI-LABEL: name: test_load_flat_v2p1_align16
; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>)) ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; VI-LABEL: name: test_load_flat_v2p1_align16 ; VI-LABEL: name: test_load_flat_v2p1_align16
; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>)) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX9-LABEL: name: test_load_flat_v2p1_align16 ; GFX9-LABEL: name: test_load_flat_v2p1_align16
; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>)) ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p0) = COPY $vgpr0_vgpr1 %0:_(p0) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 16, addrspace 0) %1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 16, addrspace 0)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
name: test_load_flat_v2p1_align8 name: test_load_flat_v2p1_align8
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_flat_v2p1_align8 ; CI-LABEL: name: test_load_flat_v2p1_align8
; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>), align 8) ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; VI-LABEL: name: test_load_flat_v2p1_align8 ; VI-LABEL: name: test_load_flat_v2p1_align8
; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>), align 8) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX9-LABEL: name: test_load_flat_v2p1_align8 ; GFX9-LABEL: name: test_load_flat_v2p1_align8
; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>), align 8) ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p0) = COPY $vgpr0_vgpr1 %0:_(p0) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 8, addrspace 0) %1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 8, addrspace 0)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
name: test_load_flat_v2p1_align4 name: test_load_flat_v2p1_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_flat_v2p1_align4 ; CI-LABEL: name: test_load_flat_v2p1_align4
; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>), align 4) ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; VI-LABEL: name: test_load_flat_v2p1_align4 ; VI-LABEL: name: test_load_flat_v2p1_align4
; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>), align 4) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX9-LABEL: name: test_load_flat_v2p1_align4 ; GFX9-LABEL: name: test_load_flat_v2p1_align4
; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>), align 4) ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p0) = COPY $vgpr0_vgpr1 %0:_(p0) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 4, addrspace 0) %1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 4, addrspace 0)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
▲ Show 20 Lines • Show All 756 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 1,634 Lines • ▼ Show 20 Lines bb.0:
; SI-LABEL: name: test_load_global_s96_align16 ; SI-LABEL: name: test_load_global_s96_align16
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s32>) = G_EXTRACT [[LOAD]](<4 x s32>), 0 ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s32>) = G_EXTRACT [[LOAD]](<4 x s32>), 0
; SI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[EXTRACT]](<3 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[EXTRACT]](<3 x s32>)
; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; CI-HSA-LABEL: name: test_load_global_s96_align16 ; CI-HSA-LABEL: name: test_load_global_s96_align16
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; CI-MESA-LABEL: name: test_load_global_s96_align16 ; CI-MESA-LABEL: name: test_load_global_s96_align16
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; VI-LABEL: name: test_load_global_s96_align16 ; VI-LABEL: name: test_load_global_s96_align16
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-HSA-LABEL: name: test_load_global_s96_align16 ; GFX9-HSA-LABEL: name: test_load_global_s96_align16
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-MESA-LABEL: name: test_load_global_s96_align16 ; GFX9-MESA-LABEL: name: test_load_global_s96_align16
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 16, addrspace 1) %1:_(s96) = G_LOAD %0 :: (load (s96), align 16, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1 $vgpr0_vgpr1_vgpr2 = COPY %1
... ...
--- ---
Show All 9 Lines bb.0:
; SI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; SI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; SI: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from unknown-address + 8, align 8, addrspace 1) ; SI: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from unknown-address + 8, align 8, addrspace 1)
; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32)
; SI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; CI-HSA-LABEL: name: test_load_global_s96_align8 ; CI-HSA-LABEL: name: test_load_global_s96_align8
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 8, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; CI-MESA-LABEL: name: test_load_global_s96_align8 ; CI-MESA-LABEL: name: test_load_global_s96_align8
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 8, addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; VI-LABEL: name: test_load_global_s96_align8 ; VI-LABEL: name: test_load_global_s96_align8
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 8, addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-HSA-LABEL: name: test_load_global_s96_align8 ; GFX9-HSA-LABEL: name: test_load_global_s96_align8
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 8, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-MESA-LABEL: name: test_load_global_s96_align8 ; GFX9-MESA-LABEL: name: test_load_global_s96_align8
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 8, addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 8, addrspace 1) %1:_(s96) = G_LOAD %0 :: (load (s96), align 8, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1 $vgpr0_vgpr1_vgpr2 = COPY %1
... ...
--- ---
Show All 9 Lines bb.0:
; SI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; SI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; SI: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from unknown-address + 8, addrspace 1) ; SI: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from unknown-address + 8, addrspace 1)
; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32)
; SI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; CI-HSA-LABEL: name: test_load_global_s96_align4 ; CI-HSA-LABEL: name: test_load_global_s96_align4
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; CI-MESA-LABEL: name: test_load_global_s96_align4 ; CI-MESA-LABEL: name: test_load_global_s96_align4
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; VI-LABEL: name: test_load_global_s96_align4 ; VI-LABEL: name: test_load_global_s96_align4
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-HSA-LABEL: name: test_load_global_s96_align4 ; GFX9-HSA-LABEL: name: test_load_global_s96_align4
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-MESA-LABEL: name: test_load_global_s96_align4 ; GFX9-MESA-LABEL: name: test_load_global_s96_align4
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 4, addrspace 1) %1:_(s96) = G_LOAD %0 :: (load (s96), align 4, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1 $vgpr0_vgpr1_vgpr2 = COPY %1
... ...
--- ---
Show All 38 Lines bb.0:
; SI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C1]] ; SI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C1]]
; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C2]](s32) ; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C2]](s32)
; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]] ; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
; SI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; CI-HSA-LABEL: name: test_load_global_s96_align2 ; CI-HSA-LABEL: name: test_load_global_s96_align2
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 2, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; CI-MESA-LABEL: name: test_load_global_s96_align2 ; CI-MESA-LABEL: name: test_load_global_s96_align2
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2 ; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; CI-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1) ; CI-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
▲ Show 20 Lines • Show All 66 Lines • ▼ Show 20 Lines bb.0:
; VI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C1]] ; VI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C1]]
; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C2]](s32) ; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C2]](s32)
; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]] ; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-HSA-LABEL: name: test_load_global_s96_align2 ; GFX9-HSA-LABEL: name: test_load_global_s96_align2
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 2, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-MESA-LABEL: name: test_load_global_s96_align2 ; GFX9-MESA-LABEL: name: test_load_global_s96_align2
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2 ; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1) ; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
▲ Show 20 Lines • Show All 117 Lines • ▼ Show 20 Lines bb.0:
; SI: [[AND11:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C3]] ; SI: [[AND11:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C3]]
; SI: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C6]](s32) ; SI: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C6]](s32)
; SI: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]] ; SI: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]]
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
; SI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; CI-HSA-LABEL: name: test_load_global_s96_align1 ; CI-HSA-LABEL: name: test_load_global_s96_align1
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 1, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; CI-MESA-LABEL: name: test_load_global_s96_align1 ; CI-MESA-LABEL: name: test_load_global_s96_align1
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1 ; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; CI-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1) ; CI-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
▲ Show 20 Lines • Show All 146 Lines • ▼ Show 20 Lines bb.0:
; VI: [[AND11:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C3]] ; VI: [[AND11:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C3]]
; VI: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C6]](s32) ; VI: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C6]](s32)
; VI: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]] ; VI: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]]
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-HSA-LABEL: name: test_load_global_s96_align1 ; GFX9-HSA-LABEL: name: test_load_global_s96_align1
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 1, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-MESA-LABEL: name: test_load_global_s96_align1 ; GFX9-MESA-LABEL: name: test_load_global_s96_align1
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1 ; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1) ; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
▲ Show 20 Lines • Show All 285 Lines • ▼ Show 20 Lines
--- ---
name: test_load_global_s128_align16 name: test_load_global_s128_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_s128_align16 ; SI-LABEL: name: test_load_global_s128_align16
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; SI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; CI-HSA-LABEL: name: test_load_global_s128_align16 ; CI-HSA-LABEL: name: test_load_global_s128_align16
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; CI-MESA-LABEL: name: test_load_global_s128_align16 ; CI-MESA-LABEL: name: test_load_global_s128_align16
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; VI-LABEL: name: test_load_global_s128_align16 ; VI-LABEL: name: test_load_global_s128_align16
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX9-HSA-LABEL: name: test_load_global_s128_align16 ; GFX9-HSA-LABEL: name: test_load_global_s128_align16
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX9-MESA-LABEL: name: test_load_global_s128_align16 ; GFX9-MESA-LABEL: name: test_load_global_s128_align16
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = G_LOAD %0 :: (load (s128), align 16, addrspace 1) %1:_(s128) = G_LOAD %0 :: (load (s128), align 16, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
name: test_load_global_s128_align4 name: test_load_global_s128_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; CI-LABEL: name: test_load_global_s128_align4 ; CI-LABEL: name: test_load_global_s128_align4
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[LOAD:%[0-9]+]]:_(s128) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1) ; CI: [[LOAD:%[0-9]+]]:_(s128) = G_LOAD [[COPY]](p1) :: (load 16, align 4, addrspace 1)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128)
; SI-LABEL: name: test_load_global_s128_align4 ; SI-LABEL: name: test_load_global_s128_align4
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
; SI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; CI-HSA-LABEL: name: test_load_global_s128_align4 ; CI-HSA-LABEL: name: test_load_global_s128_align4
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; CI-MESA-LABEL: name: test_load_global_s128_align4 ; CI-MESA-LABEL: name: test_load_global_s128_align4
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; VI-LABEL: name: test_load_global_s128_align4 ; VI-LABEL: name: test_load_global_s128_align4
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX9-HSA-LABEL: name: test_load_global_s128_align4 ; GFX9-HSA-LABEL: name: test_load_global_s128_align4
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX9-MESA-LABEL: name: test_load_global_s128_align4 ; GFX9-MESA-LABEL: name: test_load_global_s128_align4
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = G_LOAD %0 :: (load (s128), align 4, addrspace 1) %1:_(s128) = G_LOAD %0 :: (load (s128), align 4, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
▲ Show 20 Lines • Show All 101 Lines • ▼ Show 20 Lines bb.0:
; SI: [[AND15:%[0-9]+]]:_(s32) = G_AND [[COPY16]], [[C3]] ; SI: [[AND15:%[0-9]+]]:_(s32) = G_AND [[COPY16]], [[C3]]
; SI: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C6]](s32) ; SI: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C6]](s32)
; SI: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]] ; SI: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]]
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
; SI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; CI-HSA-LABEL: name: test_load_global_s128_align1 ; CI-HSA-LABEL: name: test_load_global_s128_align1
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), align 1, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 1, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; CI-MESA-LABEL: name: test_load_global_s128_align1 ; CI-MESA-LABEL: name: test_load_global_s128_align1
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1 ; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; CI-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1) ; CI-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
▲ Show 20 Lines • Show All 192 Lines • ▼ Show 20 Lines bb.0:
; VI: [[AND15:%[0-9]+]]:_(s32) = G_AND [[COPY16]], [[C3]] ; VI: [[AND15:%[0-9]+]]:_(s32) = G_AND [[COPY16]], [[C3]]
; VI: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C6]](s32) ; VI: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C6]](s32)
; VI: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]] ; VI: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]]
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX9-HSA-LABEL: name: test_load_global_s128_align1 ; GFX9-HSA-LABEL: name: test_load_global_s128_align1
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), align 1, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 1, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX9-MESA-LABEL: name: test_load_global_s128_align1 ; GFX9-MESA-LABEL: name: test_load_global_s128_align1
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1 ; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1) ; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
▲ Show 20 Lines • Show All 101 Lines • ▼ Show 20 Lines
--- ---
name: test_load_global_s256_align32 name: test_load_global_s256_align32
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_s256_align32 ; SI-LABEL: name: test_load_global_s256_align32
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (s256), align 16, addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1)
; SI: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
; SI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
; CI-HSA-LABEL: name: test_load_global_s256_align32 ; CI-HSA-LABEL: name: test_load_global_s256_align32
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (s256), align 16, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
; CI-MESA-LABEL: name: test_load_global_s256_align32 ; CI-MESA-LABEL: name: test_load_global_s256_align32
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (s256), align 16, addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
; VI-LABEL: name: test_load_global_s256_align32 ; VI-LABEL: name: test_load_global_s256_align32
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (s256), align 16, addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
; GFX9-HSA-LABEL: name: test_load_global_s256_align32 ; GFX9-HSA-LABEL: name: test_load_global_s256_align32
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (s256), align 16, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
; GFX9-MESA-LABEL: name: test_load_global_s256_align32 ; GFX9-MESA-LABEL: name: test_load_global_s256_align32
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (s256), align 16, addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s256) = G_LOAD %0 :: (load (s256), align 16, addrspace 1) %1:_(s256) = G_LOAD %0 :: (load (s256), align 16, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
... ...
--- ---
▲ Show 20 Lines • Show All 1,132 Lines • ▼ Show 20 Lines
--- ---
name: test_load_global_v2s8_align4 name: test_load_global_v2s8_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v2s8_align4 ; SI-LABEL: name: test_load_global_v2s8_align4
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), align 4, addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; SI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; SI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; SI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; SI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32) ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
; SI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; SI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]] ; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]]
; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32) ; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; SI: $vgpr0 = COPY [[ANYEXT]](s32) ; SI: $vgpr0 = COPY [[ANYEXT]](s32)
; CI-HSA-LABEL: name: test_load_global_v2s8_align4 ; CI-HSA-LABEL: name: test_load_global_v2s8_align4
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), align 4, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI-HSA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; CI-HSA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; CI-HSA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; CI-HSA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; CI-HSA: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; CI-HSA: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32) ; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
; CI-HSA: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; CI-HSA: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; CI-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI-HSA: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]] ; CI-HSA: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]]
; CI-HSA: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32) ; CI-HSA: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
; CI-HSA: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; CI-HSA: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; CI-HSA: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; CI-HSA: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; CI-HSA: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; CI-HSA: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; CI-HSA: $vgpr0 = COPY [[ANYEXT]](s32) ; CI-HSA: $vgpr0 = COPY [[ANYEXT]](s32)
; CI-MESA-LABEL: name: test_load_global_v2s8_align4 ; CI-MESA-LABEL: name: test_load_global_v2s8_align4
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), align 4, addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI-MESA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; CI-MESA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; CI-MESA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; CI-MESA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; CI-MESA: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; CI-MESA: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; CI-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32) ; CI-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
; CI-MESA: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; CI-MESA: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; CI-MESA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI-MESA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI-MESA: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]] ; CI-MESA: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]]
; CI-MESA: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32) ; CI-MESA: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
; CI-MESA: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; CI-MESA: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; CI-MESA: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; CI-MESA: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; CI-MESA: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; CI-MESA: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; CI-MESA: $vgpr0 = COPY [[ANYEXT]](s32) ; CI-MESA: $vgpr0 = COPY [[ANYEXT]](s32)
; VI-LABEL: name: test_load_global_v2s8_align4 ; VI-LABEL: name: test_load_global_v2s8_align4
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), align 4, addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; VI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; VI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]] ; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
; VI: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 ; VI: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16) ; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; VI: $vgpr0 = COPY [[ANYEXT]](s32)
; GFX9-HSA-LABEL: name: test_load_global_v2s8_align4 ; GFX9-HSA-LABEL: name: test_load_global_v2s8_align4
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), align 4, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9-HSA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; GFX9-HSA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; GFX9-HSA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; GFX9-HSA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; GFX9-HSA: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; GFX9-HSA: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; GFX9-HSA: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX9-HSA: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX9-HSA: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]] ; GFX9-HSA: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
; GFX9-HSA: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 ; GFX9-HSA: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; GFX9-HSA: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16) ; GFX9-HSA: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
; GFX9-HSA: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; GFX9-HSA: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; GFX9-HSA: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; GFX9-HSA: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; GFX9-HSA: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-HSA: $vgpr0 = COPY [[ANYEXT]](s32)
; GFX9-MESA-LABEL: name: test_load_global_v2s8_align4 ; GFX9-MESA-LABEL: name: test_load_global_v2s8_align4
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), align 4, addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9-MESA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; GFX9-MESA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; GFX9-MESA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; GFX9-MESA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
Show All 15 Lines
--- ---
name: test_load_global_v2s8_align2 name: test_load_global_v2s8_align2
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v2s8_align2 ; SI-LABEL: name: test_load_global_v2s8_align2
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; SI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; SI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; SI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; SI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32) ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
; SI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; SI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]] ; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]]
; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32) ; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; SI: $vgpr0 = COPY [[ANYEXT]](s32) ; SI: $vgpr0 = COPY [[ANYEXT]](s32)
; CI-HSA-LABEL: name: test_load_global_v2s8_align2 ; CI-HSA-LABEL: name: test_load_global_v2s8_align2
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI-HSA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; CI-HSA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; CI-HSA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; CI-HSA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; CI-HSA: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; CI-HSA: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32) ; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
; CI-HSA: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; CI-HSA: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; CI-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI-HSA: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]] ; CI-HSA: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]]
; CI-HSA: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32) ; CI-HSA: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
; CI-HSA: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; CI-HSA: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; CI-HSA: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; CI-HSA: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; CI-HSA: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; CI-HSA: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; CI-HSA: $vgpr0 = COPY [[ANYEXT]](s32) ; CI-HSA: $vgpr0 = COPY [[ANYEXT]](s32)
; CI-MESA-LABEL: name: test_load_global_v2s8_align2 ; CI-MESA-LABEL: name: test_load_global_v2s8_align2
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI-MESA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; CI-MESA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; CI-MESA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; CI-MESA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; CI-MESA: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; CI-MESA: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; CI-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32) ; CI-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
; CI-MESA: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; CI-MESA: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; CI-MESA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI-MESA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI-MESA: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]] ; CI-MESA: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]]
; CI-MESA: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32) ; CI-MESA: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
; CI-MESA: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; CI-MESA: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; CI-MESA: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; CI-MESA: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; CI-MESA: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; CI-MESA: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; CI-MESA: $vgpr0 = COPY [[ANYEXT]](s32) ; CI-MESA: $vgpr0 = COPY [[ANYEXT]](s32)
; VI-LABEL: name: test_load_global_v2s8_align2 ; VI-LABEL: name: test_load_global_v2s8_align2
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; VI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; VI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]] ; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
; VI: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 ; VI: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16) ; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; VI: $vgpr0 = COPY [[ANYEXT]](s32)
; GFX9-HSA-LABEL: name: test_load_global_v2s8_align2 ; GFX9-HSA-LABEL: name: test_load_global_v2s8_align2
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9-HSA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; GFX9-HSA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; GFX9-HSA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; GFX9-HSA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; GFX9-HSA: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; GFX9-HSA: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; GFX9-HSA: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX9-HSA: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX9-HSA: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]] ; GFX9-HSA: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
; GFX9-HSA: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 ; GFX9-HSA: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; GFX9-HSA: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16) ; GFX9-HSA: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
; GFX9-HSA: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; GFX9-HSA: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; GFX9-HSA: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; GFX9-HSA: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; GFX9-HSA: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-HSA: $vgpr0 = COPY [[ANYEXT]](s32)
; GFX9-MESA-LABEL: name: test_load_global_v2s8_align2 ; GFX9-MESA-LABEL: name: test_load_global_v2s8_align2
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9-MESA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; GFX9-MESA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; GFX9-MESA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; GFX9-MESA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
Show All 33 Lines bb.0:
; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C3]] ; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C3]]
; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C2]](s32) ; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C2]](s32)
; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; SI: $vgpr0 = COPY [[ANYEXT]](s32) ; SI: $vgpr0 = COPY [[ANYEXT]](s32)
; CI-HSA-LABEL: name: test_load_global_v2s8_align1 ; CI-HSA-LABEL: name: test_load_global_v2s8_align1
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), align 1, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI-HSA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; CI-HSA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; CI-HSA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; CI-HSA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
Show All 38 Lines bb.0:
; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C1]] ; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C1]]
; VI: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 ; VI: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C2]](s16) ; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C2]](s16)
; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; VI: $vgpr0 = COPY [[ANYEXT]](s32)
; GFX9-HSA-LABEL: name: test_load_global_v2s8_align1 ; GFX9-HSA-LABEL: name: test_load_global_v2s8_align1
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), align 1, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9-HSA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; GFX9-HSA: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; GFX9-HSA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; GFX9-HSA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
▲ Show 20 Lines • Show All 181 Lines • ▼ Show 20 Lines bb.0:
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
; SI: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[BUILD_VECTOR]](<3 x s32>) ; SI: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[BUILD_VECTOR]](<3 x s32>)
; SI: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF ; SI: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
; SI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF]](<4 x s32>) ; SI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF]](<4 x s32>)
; SI: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC1]], [[TRUNC]](<3 x s8>), 0 ; SI: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC1]], [[TRUNC]](<3 x s8>), 0
; SI: $vgpr0 = COPY [[INSERT]](<4 x s8>) ; SI: $vgpr0 = COPY [[INSERT]](<4 x s8>)
; CI-HSA-LABEL: name: test_load_global_v3s8_align1 ; CI-HSA-LABEL: name: test_load_global_v3s8_align1
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(s24) = G_LOAD [[COPY]](p1) :: (load (<3 x s8>), align 1, addrspace 1) ; CI-HSA: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(<3 x s8>) = G_BITCAST [[LOAD]](s24) ; CI-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
; CI-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-HSA: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
; CI-HSA: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
; CI-HSA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C3]](s32)
; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[OR]](s32)
; CI-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI-HSA: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; CI-HSA: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
; CI-HSA: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[BUILD_VECTOR]](<3 x s32>)
; CI-HSA: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF ; CI-HSA: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
; CI-HSA: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF]](<4 x s32>) ; CI-HSA: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF]](<4 x s32>)
; CI-HSA: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC]], [[BITCAST]](<3 x s8>), 0 ; CI-HSA: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC1]], [[TRUNC]](<3 x s8>), 0
; CI-HSA: $vgpr0 = COPY [[INSERT]](<4 x s8>) ; CI-HSA: $vgpr0 = COPY [[INSERT]](<4 x s8>)
; CI-MESA-LABEL: name: test_load_global_v3s8_align1 ; CI-MESA-LABEL: name: test_load_global_v3s8_align1
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1 ; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; CI-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1) ; CI-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
; CI-MESA: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2 ; CI-MESA: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
Show All 23 Lines bb.0:
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
; VI: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[BUILD_VECTOR]](<3 x s32>) ; VI: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[BUILD_VECTOR]](<3 x s32>)
; VI: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF ; VI: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
; VI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF]](<4 x s32>) ; VI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF]](<4 x s32>)
; VI: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC1]], [[TRUNC]](<3 x s8>), 0 ; VI: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC1]], [[TRUNC]](<3 x s8>), 0
; VI: $vgpr0 = COPY [[INSERT]](<4 x s8>) ; VI: $vgpr0 = COPY [[INSERT]](<4 x s8>)
; GFX9-HSA-LABEL: name: test_load_global_v3s8_align1 ; GFX9-HSA-LABEL: name: test_load_global_v3s8_align1
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s24) = G_LOAD [[COPY]](p1) :: (load (<3 x s8>), align 1, addrspace 1) ; GFX9-HSA: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<3 x s8>) = G_BITCAST [[LOAD]](s24) ; GFX9-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
; GFX9-HSA: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF ; GFX9-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-HSA: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF]](<4 x s32>) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
; GFX9-HSA: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC]], [[BITCAST]](<3 x s8>), 0 ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-HSA: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
; GFX9-HSA: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
; GFX9-HSA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C3]](s32)
; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[OR]](s32)
; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; GFX9-HSA: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY1]](s32), [[COPY2]](s32)
; GFX9-HSA: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; GFX9-HSA: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
; GFX9-HSA: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[DEF]](s32)
; GFX9-HSA: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
; GFX9-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[DEF1]](<2 x s16>)
; GFX9-HSA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
; GFX9-HSA: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[UV]](<3 x s16>)
; GFX9-HSA: [[DEF2:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
; GFX9-HSA: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF2]](<4 x s32>)
; GFX9-HSA: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC1]], [[TRUNC]](<3 x s8>), 0
; GFX9-HSA: $vgpr0 = COPY [[INSERT]](<4 x s8>) ; GFX9-HSA: $vgpr0 = COPY [[INSERT]](<4 x s8>)
; GFX9-MESA-LABEL: name: test_load_global_v3s8_align1 ; GFX9-MESA-LABEL: name: test_load_global_v3s8_align1
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1 ; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1) ; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
; GFX9-MESA: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2 ; GFX9-MESA: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
Show All 23 Lines
--- ---
name: test_load_global_v4s8_align4 name: test_load_global_v4s8_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v4s8_align4 ; SI-LABEL: name: test_load_global_v4s8_align4
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; SI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; SI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; SI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; SI: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; CI-HSA-LABEL: name: test_load_global_v4s8_align4 ; CI-HSA-LABEL: name: test_load_global_v4s8_align4
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; CI-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI-HSA: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; CI-HSA: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; CI-HSA: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; CI-HSA: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; CI-HSA: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; CI-HSA: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; CI-HSA: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; CI-HSA: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; CI-HSA: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; CI-HSA: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; CI-MESA-LABEL: name: test_load_global_v4s8_align4 ; CI-MESA-LABEL: name: test_load_global_v4s8_align4
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; CI-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; CI-MESA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI-MESA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI-MESA: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; CI-MESA: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; CI-MESA: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; CI-MESA: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; CI-MESA: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; CI-MESA: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; CI-MESA: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; CI-MESA: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; CI-MESA: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; CI-MESA: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; VI-LABEL: name: test_load_global_v4s8_align4 ; VI-LABEL: name: test_load_global_v4s8_align4
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; VI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; VI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; GFX9-HSA-LABEL: name: test_load_global_v4s8_align4 ; GFX9-HSA-LABEL: name: test_load_global_v4s8_align4
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; GFX9-HSA: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY1]](s32), [[COPY2]](s32) ; GFX9-HSA: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY1]](s32), [[COPY2]](s32)
; GFX9-HSA: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; GFX9-HSA: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; GFX9-HSA: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; GFX9-HSA: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; GFX9-HSA: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[COPY4]](s32) ; GFX9-HSA: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[COPY4]](s32)
; GFX9-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) ; GFX9-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
; GFX9-HSA: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>) ; GFX9-HSA: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
; GFX9-HSA: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; GFX9-HSA: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; GFX9-MESA-LABEL: name: test_load_global_v4s8_align4 ; GFX9-MESA-LABEL: name: test_load_global_v4s8_align4
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; GFX9-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; GFX9-MESA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX9-MESA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
▲ Show 20 Lines • Show All 53 Lines • ▼ Show 20 Lines bb.0:
; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LOAD1]](s32) ; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LOAD1]](s32)
; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32) ; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32)
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; SI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; SI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; SI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; SI: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; CI-HSA-LABEL: name: test_load_global_v4s8_align2 ; CI-HSA-LABEL: name: test_load_global_v4s8_align2
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), align 2, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 2, addrspace 1)
; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; CI-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
▲ Show 20 Lines • Show All 43 Lines • ▼ Show 20 Lines bb.0:
; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LOAD1]](s32) ; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LOAD1]](s32)
; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32) ; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32)
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; VI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; VI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; GFX9-HSA-LABEL: name: test_load_global_v4s8_align2 ; GFX9-HSA-LABEL: name: test_load_global_v4s8_align2
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), align 2, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 2, addrspace 1)
; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
▲ Show 20 Lines • Show All 55 Lines • ▼ Show 20 Lines bb.0:
; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LOAD1]](s32) ; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LOAD1]](s32)
; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LOAD2]](s32) ; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LOAD2]](s32)
; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LOAD3]](s32) ; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LOAD3]](s32)
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; SI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; SI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; SI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; SI: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; CI-HSA-LABEL: name: test_load_global_v4s8_align1 ; CI-HSA-LABEL: name: test_load_global_v4s8_align1
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), align 1, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 1, addrspace 1)
; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; CI-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
Show All 37 Lines bb.0:
; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LOAD1]](s32) ; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LOAD1]](s32)
; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LOAD2]](s32) ; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LOAD2]](s32)
; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LOAD3]](s32) ; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LOAD3]](s32)
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; VI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; VI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; GFX9-HSA-LABEL: name: test_load_global_v4s8_align1 ; GFX9-HSA-LABEL: name: test_load_global_v4s8_align1
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), align 1, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 1, addrspace 1)
; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
Show All 33 Lines
--- ---
name: test_load_global_v8s8_align8 name: test_load_global_v8s8_align8
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v8s8_align8 ; SI-LABEL: name: test_load_global_v8s8_align8
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s8>), addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 10 Lines bb.0:
; SI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; SI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; SI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; SI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; SI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32) ; SI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32)
; SI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>) ; SI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>)
; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>) ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>)
; SI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; SI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>)
; CI-HSA-LABEL: name: test_load_global_v8s8_align8 ; CI-HSA-LABEL: name: test_load_global_v8s8_align8
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s8>), addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
; CI-HSA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; CI-HSA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 10 Lines bb.0:
; CI-HSA: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; CI-HSA: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; CI-HSA: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; CI-HSA: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; CI-HSA: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32) ; CI-HSA: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32)
; CI-HSA: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>) ; CI-HSA: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>)
; CI-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>) ; CI-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>)
; CI-HSA: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; CI-HSA: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>)
; CI-MESA-LABEL: name: test_load_global_v8s8_align8 ; CI-MESA-LABEL: name: test_load_global_v8s8_align8
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s8>), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
; CI-MESA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; CI-MESA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; CI-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; CI-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; CI-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; CI-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; CI-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 10 Lines bb.0:
; CI-MESA: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; CI-MESA: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; CI-MESA: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; CI-MESA: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; CI-MESA: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32) ; CI-MESA: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32)
; CI-MESA: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>) ; CI-MESA: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>)
; CI-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>) ; CI-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>)
; CI-MESA: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; CI-MESA: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>)
; VI-LABEL: name: test_load_global_v8s8_align8 ; VI-LABEL: name: test_load_global_v8s8_align8
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s8>), addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 10 Lines bb.0:
; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; VI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; VI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; VI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32) ; VI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32)
; VI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>) ; VI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>)
; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>) ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>)
; VI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; VI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>)
; GFX9-HSA-LABEL: name: test_load_global_v8s8_align8 ; GFX9-HSA-LABEL: name: test_load_global_v8s8_align8
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s8>), addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
; GFX9-HSA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX9-HSA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 14 Lines bb.0:
; GFX9-HSA: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; GFX9-HSA: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; GFX9-HSA: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY7]](s32), [[COPY8]](s32) ; GFX9-HSA: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY7]](s32), [[COPY8]](s32)
; GFX9-HSA: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>) ; GFX9-HSA: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>)
; GFX9-HSA: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS1]](<4 x s16>) ; GFX9-HSA: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS1]](<4 x s16>)
; GFX9-HSA: [[CONCAT_VECTORS2:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>) ; GFX9-HSA: [[CONCAT_VECTORS2:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>)
; GFX9-HSA: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS2]](<8 x s8>) ; GFX9-HSA: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS2]](<8 x s8>)
; GFX9-MESA-LABEL: name: test_load_global_v8s8_align8 ; GFX9-MESA-LABEL: name: test_load_global_v8s8_align8
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s8>), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
; GFX9-MESA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX9-MESA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; GFX9-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; GFX9-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; GFX9-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; GFX9-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; GFX9-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 25 Lines
--- ---
name: test_load_global_v16s8_align16 name: test_load_global_v16s8_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v16s8_align16 ; SI-LABEL: name: test_load_global_v16s8_align16
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s8>), addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 28 Lines bb.0:
; SI: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32) ; SI: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32)
; SI: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32) ; SI: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32)
; SI: [[BUILD_VECTOR3:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32) ; SI: [[BUILD_VECTOR3:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32)
; SI: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR3]](<4 x s32>) ; SI: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR3]](<4 x s32>)
; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>) ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>)
; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>)
; CI-HSA-LABEL: name: test_load_global_v16s8_align16 ; CI-HSA-LABEL: name: test_load_global_v16s8_align16
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s8>), addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; CI-HSA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; CI-HSA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 28 Lines bb.0:
; CI-HSA: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32) ; CI-HSA: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32)
; CI-HSA: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32) ; CI-HSA: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32)
; CI-HSA: [[BUILD_VECTOR3:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32) ; CI-HSA: [[BUILD_VECTOR3:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32)
; CI-HSA: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR3]](<4 x s32>) ; CI-HSA: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR3]](<4 x s32>)
; CI-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>) ; CI-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>)
; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>)
; CI-MESA-LABEL: name: test_load_global_v16s8_align16 ; CI-MESA-LABEL: name: test_load_global_v16s8_align16
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s8>), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; CI-MESA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; CI-MESA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; CI-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; CI-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; CI-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; CI-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; CI-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 28 Lines bb.0:
; CI-MESA: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32) ; CI-MESA: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32)
; CI-MESA: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32) ; CI-MESA: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32)
; CI-MESA: [[BUILD_VECTOR3:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32) ; CI-MESA: [[BUILD_VECTOR3:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32)
; CI-MESA: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR3]](<4 x s32>) ; CI-MESA: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR3]](<4 x s32>)
; CI-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>) ; CI-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>)
; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>)
; VI-LABEL: name: test_load_global_v16s8_align16 ; VI-LABEL: name: test_load_global_v16s8_align16
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s8>), addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 28 Lines bb.0:
; VI: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32) ; VI: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32)
; VI: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32) ; VI: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32)
; VI: [[BUILD_VECTOR3:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32) ; VI: [[BUILD_VECTOR3:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32)
; VI: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR3]](<4 x s32>) ; VI: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR3]](<4 x s32>)
; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>) ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>)
; GFX9-HSA-LABEL: name: test_load_global_v16s8_align16 ; GFX9-HSA-LABEL: name: test_load_global_v16s8_align16
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s8>), addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-HSA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; GFX9-HSA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 36 Lines bb.0:
; GFX9-HSA: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32) ; GFX9-HSA: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32)
; GFX9-HSA: [[BUILD_VECTOR_TRUNC7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY15]](s32), [[COPY16]](s32) ; GFX9-HSA: [[BUILD_VECTOR_TRUNC7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY15]](s32), [[COPY16]](s32)
; GFX9-HSA: [[CONCAT_VECTORS3:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC6]](<2 x s16>), [[BUILD_VECTOR_TRUNC7]](<2 x s16>) ; GFX9-HSA: [[CONCAT_VECTORS3:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC6]](<2 x s16>), [[BUILD_VECTOR_TRUNC7]](<2 x s16>)
; GFX9-HSA: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS3]](<4 x s16>) ; GFX9-HSA: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS3]](<4 x s16>)
; GFX9-HSA: [[CONCAT_VECTORS4:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>) ; GFX9-HSA: [[CONCAT_VECTORS4:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS4]](<16 x s8>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS4]](<16 x s8>)
; GFX9-MESA-LABEL: name: test_load_global_v16s8_align16 ; GFX9-MESA-LABEL: name: test_load_global_v16s8_align16
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s8>), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-MESA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; GFX9-MESA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; GFX9-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; GFX9-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; GFX9-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; GFX9-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; GFX9-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
▲ Show 20 Lines • Show All 47 Lines • ▼ Show 20 Lines
--- ---
name: test_load_global_v32s8_align32 name: test_load_global_v32s8_align32
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v32s8_align32 ; SI-LABEL: name: test_load_global_v32s8_align32
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<32 x s8>), addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>)
; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
▲ Show 20 Lines • Show All 64 Lines • ▼ Show 20 Lines bb.0:
; SI: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR22]](s32) ; SI: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR22]](s32)
; SI: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR23]](s32) ; SI: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR23]](s32)
; SI: [[BUILD_VECTOR7:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32) ; SI: [[BUILD_VECTOR7:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32)
; SI: [[TRUNC7:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR7]](<4 x s32>) ; SI: [[TRUNC7:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR7]](<4 x s32>)
; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>), [[TRUNC4]](<4 x s8>), [[TRUNC5]](<4 x s8>), [[TRUNC6]](<4 x s8>), [[TRUNC7]](<4 x s8>) ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>), [[TRUNC4]](<4 x s8>), [[TRUNC5]](<4 x s8>), [[TRUNC6]](<4 x s8>), [[TRUNC7]](<4 x s8>)
; SI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>)
; CI-HSA-LABEL: name: test_load_global_v32s8_align32 ; CI-HSA-LABEL: name: test_load_global_v32s8_align32
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<32 x s8>), addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
; CI-HSA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; CI-HSA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>)
; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
▲ Show 20 Lines • Show All 64 Lines • ▼ Show 20 Lines bb.0:
; CI-HSA: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR22]](s32) ; CI-HSA: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR22]](s32)
; CI-HSA: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR23]](s32) ; CI-HSA: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR23]](s32)
; CI-HSA: [[BUILD_VECTOR7:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32) ; CI-HSA: [[BUILD_VECTOR7:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32)
; CI-HSA: [[TRUNC7:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR7]](<4 x s32>) ; CI-HSA: [[TRUNC7:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR7]](<4 x s32>)
; CI-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>), [[TRUNC4]](<4 x s8>), [[TRUNC5]](<4 x s8>), [[TRUNC6]](<4 x s8>), [[TRUNC7]](<4 x s8>) ; CI-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>), [[TRUNC4]](<4 x s8>), [[TRUNC5]](<4 x s8>), [[TRUNC6]](<4 x s8>), [[TRUNC7]](<4 x s8>)
; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>)
; CI-MESA-LABEL: name: test_load_global_v32s8_align32 ; CI-MESA-LABEL: name: test_load_global_v32s8_align32
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<32 x s8>), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
; CI-MESA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; CI-MESA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>)
; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; CI-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; CI-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; CI-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; CI-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; CI-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
▲ Show 20 Lines • Show All 64 Lines • ▼ Show 20 Lines bb.0:
; CI-MESA: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR22]](s32) ; CI-MESA: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR22]](s32)
; CI-MESA: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR23]](s32) ; CI-MESA: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR23]](s32)
; CI-MESA: [[BUILD_VECTOR7:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32) ; CI-MESA: [[BUILD_VECTOR7:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32)
; CI-MESA: [[TRUNC7:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR7]](<4 x s32>) ; CI-MESA: [[TRUNC7:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR7]](<4 x s32>)
; CI-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>), [[TRUNC4]](<4 x s8>), [[TRUNC5]](<4 x s8>), [[TRUNC6]](<4 x s8>), [[TRUNC7]](<4 x s8>) ; CI-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>), [[TRUNC4]](<4 x s8>), [[TRUNC5]](<4 x s8>), [[TRUNC6]](<4 x s8>), [[TRUNC7]](<4 x s8>)
; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>)
; VI-LABEL: name: test_load_global_v32s8_align32 ; VI-LABEL: name: test_load_global_v32s8_align32
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<32 x s8>), addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
▲ Show 20 Lines • Show All 64 Lines • ▼ Show 20 Lines bb.0:
; VI: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR22]](s32) ; VI: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR22]](s32)
; VI: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR23]](s32) ; VI: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR23]](s32)
; VI: [[BUILD_VECTOR7:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32) ; VI: [[BUILD_VECTOR7:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32)
; VI: [[TRUNC7:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR7]](<4 x s32>) ; VI: [[TRUNC7:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR7]](<4 x s32>)
; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>), [[TRUNC4]](<4 x s8>), [[TRUNC5]](<4 x s8>), [[TRUNC6]](<4 x s8>), [[TRUNC7]](<4 x s8>) ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>), [[TRUNC4]](<4 x s8>), [[TRUNC5]](<4 x s8>), [[TRUNC6]](<4 x s8>), [[TRUNC7]](<4 x s8>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>)
; GFX9-HSA-LABEL: name: test_load_global_v32s8_align32 ; GFX9-HSA-LABEL: name: test_load_global_v32s8_align32
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<32 x s8>), addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
; GFX9-HSA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; GFX9-HSA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>)
; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
▲ Show 20 Lines • Show All 80 Lines • ▼ Show 20 Lines bb.0:
; GFX9-HSA: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR23]](s32) ; GFX9-HSA: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR23]](s32)
; GFX9-HSA: [[BUILD_VECTOR_TRUNC15:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY31]](s32), [[COPY32]](s32) ; GFX9-HSA: [[BUILD_VECTOR_TRUNC15:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY31]](s32), [[COPY32]](s32)
; GFX9-HSA: [[CONCAT_VECTORS7:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC14]](<2 x s16>), [[BUILD_VECTOR_TRUNC15]](<2 x s16>) ; GFX9-HSA: [[CONCAT_VECTORS7:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC14]](<2 x s16>), [[BUILD_VECTOR_TRUNC15]](<2 x s16>)
; GFX9-HSA: [[TRUNC7:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS7]](<4 x s16>) ; GFX9-HSA: [[TRUNC7:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS7]](<4 x s16>)
; GFX9-HSA: [[CONCAT_VECTORS8:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>), [[TRUNC4]](<4 x s8>), [[TRUNC5]](<4 x s8>), [[TRUNC6]](<4 x s8>), [[TRUNC7]](<4 x s8>) ; GFX9-HSA: [[CONCAT_VECTORS8:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>), [[TRUNC4]](<4 x s8>), [[TRUNC5]](<4 x s8>), [[TRUNC6]](<4 x s8>), [[TRUNC7]](<4 x s8>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS8]](<32 x s8>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS8]](<32 x s8>)
; GFX9-MESA-LABEL: name: test_load_global_v32s8_align32 ; GFX9-MESA-LABEL: name: test_load_global_v32s8_align32
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<32 x s8>), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
; GFX9-MESA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; GFX9-MESA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>)
; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; GFX9-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; GFX9-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; GFX9-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; GFX9-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; GFX9-MESA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
▲ Show 20 Lines • Show All 1,354 Lines • ▼ Show 20 Lines
--- ---
name: test_load_global_v5s16_align16 name: test_load_global_v5s16_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v5s16_align16 ; SI-LABEL: name: test_load_global_v5s16_align16
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; SI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; SI: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; SI: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; SI: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF ; SI: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
; SI: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF ; SI: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>) ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
Show All 26 Lines bb.0:
; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) ; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]] ; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
; SI: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) ; SI: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
; SI: $vgpr0 = COPY [[BITCAST5]](<2 x s16>) ; SI: $vgpr0 = COPY [[BITCAST5]](<2 x s16>)
; SI: $vgpr1 = COPY [[BITCAST6]](<2 x s16>) ; SI: $vgpr1 = COPY [[BITCAST6]](<2 x s16>)
; SI: $vgpr2 = COPY [[BITCAST7]](<2 x s16>) ; SI: $vgpr2 = COPY [[BITCAST7]](<2 x s16>)
; CI-HSA-LABEL: name: test_load_global_v5s16_align16 ; CI-HSA-LABEL: name: test_load_global_v5s16_align16
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-HSA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; CI-HSA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; CI-HSA: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF ; CI-HSA: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
; CI-HSA: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF ; CI-HSA: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
; CI-HSA: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>) ; CI-HSA: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
; CI-HSA: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; CI-HSA: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
Show All 26 Lines bb.0:
; CI-HSA: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) ; CI-HSA: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
; CI-HSA: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]] ; CI-HSA: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
; CI-HSA: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) ; CI-HSA: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
; CI-HSA: $vgpr0 = COPY [[BITCAST5]](<2 x s16>) ; CI-HSA: $vgpr0 = COPY [[BITCAST5]](<2 x s16>)
; CI-HSA: $vgpr1 = COPY [[BITCAST6]](<2 x s16>) ; CI-HSA: $vgpr1 = COPY [[BITCAST6]](<2 x s16>)
; CI-HSA: $vgpr2 = COPY [[BITCAST7]](<2 x s16>) ; CI-HSA: $vgpr2 = COPY [[BITCAST7]](<2 x s16>)
; CI-MESA-LABEL: name: test_load_global_v5s16_align16 ; CI-MESA-LABEL: name: test_load_global_v5s16_align16
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-MESA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; CI-MESA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; CI-MESA: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF ; CI-MESA: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
; CI-MESA: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF ; CI-MESA: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
; CI-MESA: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>) ; CI-MESA: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
; CI-MESA: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; CI-MESA: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
Show All 26 Lines bb.0:
; CI-MESA: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) ; CI-MESA: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
; CI-MESA: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]] ; CI-MESA: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
; CI-MESA: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) ; CI-MESA: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
; CI-MESA: $vgpr0 = COPY [[BITCAST5]](<2 x s16>) ; CI-MESA: $vgpr0 = COPY [[BITCAST5]](<2 x s16>)
; CI-MESA: $vgpr1 = COPY [[BITCAST6]](<2 x s16>) ; CI-MESA: $vgpr1 = COPY [[BITCAST6]](<2 x s16>)
; CI-MESA: $vgpr2 = COPY [[BITCAST7]](<2 x s16>) ; CI-MESA: $vgpr2 = COPY [[BITCAST7]](<2 x s16>)
; VI-LABEL: name: test_load_global_v5s16_align16 ; VI-LABEL: name: test_load_global_v5s16_align16
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; VI: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; VI: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF ; VI: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
; VI: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF ; VI: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>) ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
Show All 26 Lines bb.0:
; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) ; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]] ; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
; VI: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) ; VI: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
; VI: $vgpr0 = COPY [[BITCAST5]](<2 x s16>) ; VI: $vgpr0 = COPY [[BITCAST5]](<2 x s16>)
; VI: $vgpr1 = COPY [[BITCAST6]](<2 x s16>) ; VI: $vgpr1 = COPY [[BITCAST6]](<2 x s16>)
; VI: $vgpr2 = COPY [[BITCAST7]](<2 x s16>) ; VI: $vgpr2 = COPY [[BITCAST7]](<2 x s16>)
; GFX9-HSA-LABEL: name: test_load_global_v5s16_align16 ; GFX9-HSA-LABEL: name: test_load_global_v5s16_align16
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-HSA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; GFX9-HSA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; GFX9-HSA: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF ; GFX9-HSA: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
; GFX9-HSA: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF ; GFX9-HSA: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
; GFX9-HSA: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>) ; GFX9-HSA: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
; GFX9-HSA: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX9-HSA: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
Show All 13 Lines bb.0:
; GFX9-HSA: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) ; GFX9-HSA: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32)
; GFX9-HSA: [[COPY6:%[0-9]+]]:_(s32) = COPY [[BITCAST4]](s32) ; GFX9-HSA: [[COPY6:%[0-9]+]]:_(s32) = COPY [[BITCAST4]](s32)
; GFX9-HSA: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY5]](s32), [[COPY6]](s32) ; GFX9-HSA: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY5]](s32), [[COPY6]](s32)
; GFX9-HSA: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC]](<2 x s16>) ; GFX9-HSA: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC]](<2 x s16>)
; GFX9-HSA: $vgpr1 = COPY [[BUILD_VECTOR_TRUNC1]](<2 x s16>) ; GFX9-HSA: $vgpr1 = COPY [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
; GFX9-HSA: $vgpr2 = COPY [[BUILD_VECTOR_TRUNC2]](<2 x s16>) ; GFX9-HSA: $vgpr2 = COPY [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
; GFX9-MESA-LABEL: name: test_load_global_v5s16_align16 ; GFX9-MESA-LABEL: name: test_load_global_v5s16_align16
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; GFX9-MESA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; GFX9-MESA: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF ; GFX9-MESA: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
; GFX9-MESA: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF ; GFX9-MESA: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
; GFX9-MESA: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>) ; GFX9-MESA: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
; GFX9-MESA: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX9-MESA: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
▲ Show 20 Lines • Show All 891 Lines • ▼ Show 20 Lines bb.0:
; SI-LABEL: name: test_load_global_v6s16_align16 ; SI-LABEL: name: test_load_global_v6s16_align16
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s32>) = G_EXTRACT [[LOAD]](<4 x s32>), 0 ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s32>) = G_EXTRACT [[LOAD]](<4 x s32>), 0
; SI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[EXTRACT]](<3 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[EXTRACT]](<3 x s32>)
; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; CI-HSA-LABEL: name: test_load_global_v6s16_align16 ; CI-HSA-LABEL: name: test_load_global_v6s16_align16
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 16, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; CI-MESA-LABEL: name: test_load_global_v6s16_align16 ; CI-MESA-LABEL: name: test_load_global_v6s16_align16
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 16, addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; VI-LABEL: name: test_load_global_v6s16_align16 ; VI-LABEL: name: test_load_global_v6s16_align16
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 16, addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; GFX9-HSA-LABEL: name: test_load_global_v6s16_align16 ; GFX9-HSA-LABEL: name: test_load_global_v6s16_align16
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 16, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; GFX9-MESA-LABEL: name: test_load_global_v6s16_align16 ; GFX9-MESA-LABEL: name: test_load_global_v6s16_align16
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 16, addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<6 x s16>) = G_LOAD %0 :: (load (<6 x s16>), align 16, addrspace 1) %1:_(<6 x s16>) = G_LOAD %0 :: (load (<6 x s16>), align 16, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1 $vgpr0_vgpr1_vgpr2 = COPY %1
... ...
--- ---
Show All 9 Lines bb.0:
; SI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; SI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; SI: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from unknown-address + 8, align 8, addrspace 1) ; SI: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from unknown-address + 8, align 8, addrspace 1)
; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32)
; SI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; CI-HSA-LABEL: name: test_load_global_v6s16_align8 ; CI-HSA-LABEL: name: test_load_global_v6s16_align8
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 8, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; CI-MESA-LABEL: name: test_load_global_v6s16_align8 ; CI-MESA-LABEL: name: test_load_global_v6s16_align8
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 8, addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; VI-LABEL: name: test_load_global_v6s16_align8 ; VI-LABEL: name: test_load_global_v6s16_align8
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 8, addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; GFX9-HSA-LABEL: name: test_load_global_v6s16_align8 ; GFX9-HSA-LABEL: name: test_load_global_v6s16_align8
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 8, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; GFX9-MESA-LABEL: name: test_load_global_v6s16_align8 ; GFX9-MESA-LABEL: name: test_load_global_v6s16_align8
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 8, addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<6 x s16>) = G_LOAD %0 :: (load (<6 x s16>), align 8, addrspace 1) %1:_(<6 x s16>) = G_LOAD %0 :: (load (<6 x s16>), align 8, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1 $vgpr0_vgpr1_vgpr2 = COPY %1
... ...
--- ---
Show All 9 Lines bb.0:
; SI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; SI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; SI: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from unknown-address + 8, addrspace 1) ; SI: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from unknown-address + 8, addrspace 1)
; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32)
; SI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; CI-HSA-LABEL: name: test_load_global_v6s16_align4 ; CI-HSA-LABEL: name: test_load_global_v6s16_align4
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 4, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; CI-MESA-LABEL: name: test_load_global_v6s16_align4 ; CI-MESA-LABEL: name: test_load_global_v6s16_align4
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 4, addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; VI-LABEL: name: test_load_global_v6s16_align4 ; VI-LABEL: name: test_load_global_v6s16_align4
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 4, addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; GFX9-HSA-LABEL: name: test_load_global_v6s16_align4 ; GFX9-HSA-LABEL: name: test_load_global_v6s16_align4
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 4, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; GFX9-MESA-LABEL: name: test_load_global_v6s16_align4 ; GFX9-MESA-LABEL: name: test_load_global_v6s16_align4
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 4, addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<6 x s16>) = G_LOAD %0 :: (load (<6 x s16>), align 4, addrspace 1) %1:_(<6 x s16>) = G_LOAD %0 :: (load (<6 x s16>), align 4, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1 $vgpr0_vgpr1_vgpr2 = COPY %1
... ...
--- ---
Show All 38 Lines bb.0:
; SI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C1]] ; SI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C1]]
; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C2]](s32) ; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C2]](s32)
; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]] ; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
; SI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; CI-HSA-LABEL: name: test_load_global_v6s16_align2 ; CI-HSA-LABEL: name: test_load_global_v6s16_align2
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 2, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; CI-MESA-LABEL: name: test_load_global_v6s16_align2 ; CI-MESA-LABEL: name: test_load_global_v6s16_align2
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2 ; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; CI-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1) ; CI-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
▲ Show 20 Lines • Show All 66 Lines • ▼ Show 20 Lines bb.0:
; VI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C1]] ; VI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C1]]
; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C2]](s32) ; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C2]](s32)
; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]] ; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
; VI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; GFX9-HSA-LABEL: name: test_load_global_v6s16_align2 ; GFX9-HSA-LABEL: name: test_load_global_v6s16_align2
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 2, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; GFX9-MESA-LABEL: name: test_load_global_v6s16_align2 ; GFX9-MESA-LABEL: name: test_load_global_v6s16_align2
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2 ; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1) ; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
▲ Show 20 Lines • Show All 117 Lines • ▼ Show 20 Lines bb.0:
; SI: [[AND11:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C3]] ; SI: [[AND11:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C3]]
; SI: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C6]](s32) ; SI: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C6]](s32)
; SI: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]] ; SI: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]]
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
; SI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; CI-HSA-LABEL: name: test_load_global_v6s16_align1 ; CI-HSA-LABEL: name: test_load_global_v6s16_align1
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 1, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; CI-MESA-LABEL: name: test_load_global_v6s16_align1 ; CI-MESA-LABEL: name: test_load_global_v6s16_align1
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1 ; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; CI-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1) ; CI-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
▲ Show 20 Lines • Show All 146 Lines • ▼ Show 20 Lines bb.0:
; VI: [[AND11:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C3]] ; VI: [[AND11:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C3]]
; VI: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C6]](s32) ; VI: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C6]](s32)
; VI: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]] ; VI: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]]
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
; VI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; GFX9-HSA-LABEL: name: test_load_global_v6s16_align1 ; GFX9-HSA-LABEL: name: test_load_global_v6s16_align1
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 1, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
; GFX9-MESA-LABEL: name: test_load_global_v6s16_align1 ; GFX9-MESA-LABEL: name: test_load_global_v6s16_align1
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1 ; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1) ; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
▲ Show 20 Lines • Show All 78 Lines • ▼ Show 20 Lines
--- ---
name: test_load_global_v7s16_align16 name: test_load_global_v7s16_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v7s16_align16 ; SI-LABEL: name: test_load_global_v7s16_align16
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; SI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; SI: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; SI: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; SI: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; SI: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>) ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
Show All 35 Lines bb.0:
; SI: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]] ; SI: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
; SI: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) ; SI: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
; SI: $vgpr0 = COPY [[BITCAST6]](<2 x s16>) ; SI: $vgpr0 = COPY [[BITCAST6]](<2 x s16>)
; SI: $vgpr1 = COPY [[BITCAST7]](<2 x s16>) ; SI: $vgpr1 = COPY [[BITCAST7]](<2 x s16>)
; SI: $vgpr2 = COPY [[BITCAST8]](<2 x s16>) ; SI: $vgpr2 = COPY [[BITCAST8]](<2 x s16>)
; SI: $vgpr3 = COPY [[BITCAST9]](<2 x s16>) ; SI: $vgpr3 = COPY [[BITCAST9]](<2 x s16>)
; CI-HSA-LABEL: name: test_load_global_v7s16_align16 ; CI-HSA-LABEL: name: test_load_global_v7s16_align16
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-HSA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; CI-HSA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; CI-HSA: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; CI-HSA: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; CI-HSA: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>) ; CI-HSA: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
; CI-HSA: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; CI-HSA: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; CI-HSA: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; CI-HSA: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
Show All 35 Lines bb.0:
; CI-HSA: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]] ; CI-HSA: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
; CI-HSA: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) ; CI-HSA: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
; CI-HSA: $vgpr0 = COPY [[BITCAST6]](<2 x s16>) ; CI-HSA: $vgpr0 = COPY [[BITCAST6]](<2 x s16>)
; CI-HSA: $vgpr1 = COPY [[BITCAST7]](<2 x s16>) ; CI-HSA: $vgpr1 = COPY [[BITCAST7]](<2 x s16>)
; CI-HSA: $vgpr2 = COPY [[BITCAST8]](<2 x s16>) ; CI-HSA: $vgpr2 = COPY [[BITCAST8]](<2 x s16>)
; CI-HSA: $vgpr3 = COPY [[BITCAST9]](<2 x s16>) ; CI-HSA: $vgpr3 = COPY [[BITCAST9]](<2 x s16>)
; CI-MESA-LABEL: name: test_load_global_v7s16_align16 ; CI-MESA-LABEL: name: test_load_global_v7s16_align16
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-MESA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; CI-MESA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; CI-MESA: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; CI-MESA: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; CI-MESA: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>) ; CI-MESA: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
; CI-MESA: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; CI-MESA: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; CI-MESA: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; CI-MESA: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
Show All 35 Lines bb.0:
; CI-MESA: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]] ; CI-MESA: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
; CI-MESA: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) ; CI-MESA: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
; CI-MESA: $vgpr0 = COPY [[BITCAST6]](<2 x s16>) ; CI-MESA: $vgpr0 = COPY [[BITCAST6]](<2 x s16>)
; CI-MESA: $vgpr1 = COPY [[BITCAST7]](<2 x s16>) ; CI-MESA: $vgpr1 = COPY [[BITCAST7]](<2 x s16>)
; CI-MESA: $vgpr2 = COPY [[BITCAST8]](<2 x s16>) ; CI-MESA: $vgpr2 = COPY [[BITCAST8]](<2 x s16>)
; CI-MESA: $vgpr3 = COPY [[BITCAST9]](<2 x s16>) ; CI-MESA: $vgpr3 = COPY [[BITCAST9]](<2 x s16>)
; VI-LABEL: name: test_load_global_v7s16_align16 ; VI-LABEL: name: test_load_global_v7s16_align16
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; VI: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; VI: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; VI: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>) ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
Show All 35 Lines bb.0:
; VI: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]] ; VI: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
; VI: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) ; VI: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
; VI: $vgpr0 = COPY [[BITCAST6]](<2 x s16>) ; VI: $vgpr0 = COPY [[BITCAST6]](<2 x s16>)
; VI: $vgpr1 = COPY [[BITCAST7]](<2 x s16>) ; VI: $vgpr1 = COPY [[BITCAST7]](<2 x s16>)
; VI: $vgpr2 = COPY [[BITCAST8]](<2 x s16>) ; VI: $vgpr2 = COPY [[BITCAST8]](<2 x s16>)
; VI: $vgpr3 = COPY [[BITCAST9]](<2 x s16>) ; VI: $vgpr3 = COPY [[BITCAST9]](<2 x s16>)
; GFX9-HSA-LABEL: name: test_load_global_v7s16_align16 ; GFX9-HSA-LABEL: name: test_load_global_v7s16_align16
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-HSA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; GFX9-HSA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; GFX9-HSA: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; GFX9-HSA: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; GFX9-HSA: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>) ; GFX9-HSA: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
; GFX9-HSA: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX9-HSA: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; GFX9-HSA: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; GFX9-HSA: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
Show All 18 Lines bb.0:
; GFX9-HSA: [[COPY8:%[0-9]+]]:_(s32) = COPY [[BITCAST5]](s32) ; GFX9-HSA: [[COPY8:%[0-9]+]]:_(s32) = COPY [[BITCAST5]](s32)
; GFX9-HSA: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY7]](s32), [[COPY8]](s32) ; GFX9-HSA: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY7]](s32), [[COPY8]](s32)
; GFX9-HSA: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC]](<2 x s16>) ; GFX9-HSA: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC]](<2 x s16>)
; GFX9-HSA: $vgpr1 = COPY [[BUILD_VECTOR_TRUNC1]](<2 x s16>) ; GFX9-HSA: $vgpr1 = COPY [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
; GFX9-HSA: $vgpr2 = COPY [[BUILD_VECTOR_TRUNC2]](<2 x s16>) ; GFX9-HSA: $vgpr2 = COPY [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
; GFX9-HSA: $vgpr3 = COPY [[BUILD_VECTOR_TRUNC3]](<2 x s16>) ; GFX9-HSA: $vgpr3 = COPY [[BUILD_VECTOR_TRUNC3]](<2 x s16>)
; GFX9-MESA-LABEL: name: test_load_global_v7s16_align16 ; GFX9-MESA-LABEL: name: test_load_global_v7s16_align16
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; GFX9-MESA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; GFX9-MESA: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; GFX9-MESA: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
; GFX9-MESA: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>) ; GFX9-MESA: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
; GFX9-MESA: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) ; GFX9-MESA: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; GFX9-MESA: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) ; GFX9-MESA: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
▲ Show 20 Lines • Show All 1,674 Lines • ▼ Show 20 Lines
--- ---
name: test_load_global_v8s16_align16 name: test_load_global_v8s16_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v8s16_align16 ; SI-LABEL: name: test_load_global_v8s16_align16
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; SI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
; CI-HSA-LABEL: name: test_load_global_v8s16_align16 ; CI-HSA-LABEL: name: test_load_global_v8s16_align16
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
; CI-MESA-LABEL: name: test_load_global_v8s16_align16 ; CI-MESA-LABEL: name: test_load_global_v8s16_align16
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
; VI-LABEL: name: test_load_global_v8s16_align16 ; VI-LABEL: name: test_load_global_v8s16_align16
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
; GFX9-HSA-LABEL: name: test_load_global_v8s16_align16 ; GFX9-HSA-LABEL: name: test_load_global_v8s16_align16
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
; GFX9-MESA-LABEL: name: test_load_global_v8s16_align16 ; GFX9-MESA-LABEL: name: test_load_global_v8s16_align16
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s16>) = G_LOAD %0 :: (load (<8 x s16>), align 16, addrspace 1) %1:_(<8 x s16>) = G_LOAD %0 :: (load (<8 x s16>), align 16, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
name: test_load_global_v8s16_align8 name: test_load_global_v8s16_align8
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v8s16_align8 ; SI-LABEL: name: test_load_global_v8s16_align8
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), align 8, addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
; SI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
; CI-HSA-LABEL: name: test_load_global_v8s16_align8 ; CI-HSA-LABEL: name: test_load_global_v8s16_align8
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), align 8, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
; CI-MESA-LABEL: name: test_load_global_v8s16_align8 ; CI-MESA-LABEL: name: test_load_global_v8s16_align8
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), align 8, addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
; VI-LABEL: name: test_load_global_v8s16_align8 ; VI-LABEL: name: test_load_global_v8s16_align8
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), align 8, addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
; GFX9-HSA-LABEL: name: test_load_global_v8s16_align8 ; GFX9-HSA-LABEL: name: test_load_global_v8s16_align8
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), align 8, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
; GFX9-MESA-LABEL: name: test_load_global_v8s16_align8 ; GFX9-MESA-LABEL: name: test_load_global_v8s16_align8
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), align 8, addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s16>) = G_LOAD %0 :: (load (<8 x s16>), align 8, addrspace 1) %1:_(<8 x s16>) = G_LOAD %0 :: (load (<8 x s16>), align 8, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
▲ Show 20 Lines • Show All 1,563 Lines • ▼ Show 20 Lines
--- ---
name: test_load_global_v2sp1_align16 name: test_load_global_v2sp1_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v2sp1_align16 ; SI-LABEL: name: test_load_global_v2sp1_align16
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; SI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; CI-HSA-LABEL: name: test_load_global_v2sp1_align16 ; CI-HSA-LABEL: name: test_load_global_v2sp1_align16
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; CI-MESA-LABEL: name: test_load_global_v2sp1_align16 ; CI-MESA-LABEL: name: test_load_global_v2sp1_align16
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; VI-LABEL: name: test_load_global_v2sp1_align16 ; VI-LABEL: name: test_load_global_v2sp1_align16
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX9-HSA-LABEL: name: test_load_global_v2sp1_align16 ; GFX9-HSA-LABEL: name: test_load_global_v2sp1_align16
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX9-MESA-LABEL: name: test_load_global_v2sp1_align16 ; GFX9-MESA-LABEL: name: test_load_global_v2sp1_align16
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 16, addrspace 1) %1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 16, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
▲ Show 20 Lines • Show All 1,873 Lines • ▼ Show 20 Lines
--- ---
name: test_load_global_v2s128_align32 name: test_load_global_v2s128_align32
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v2s128_align32 ; SI-LABEL: name: test_load_global_v2s128_align32
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s128>), addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
; SI: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
; SI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
; CI-HSA-LABEL: name: test_load_global_v2s128_align32 ; CI-HSA-LABEL: name: test_load_global_v2s128_align32
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s128>), addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
; CI-MESA-LABEL: name: test_load_global_v2s128_align32 ; CI-MESA-LABEL: name: test_load_global_v2s128_align32
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s128>), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
; VI-LABEL: name: test_load_global_v2s128_align32 ; VI-LABEL: name: test_load_global_v2s128_align32
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s128>), addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
; GFX9-HSA-LABEL: name: test_load_global_v2s128_align32 ; GFX9-HSA-LABEL: name: test_load_global_v2s128_align32
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s128>), addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
; GFX9-MESA-LABEL: name: test_load_global_v2s128_align32 ; GFX9-MESA-LABEL: name: test_load_global_v2s128_align32
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s128>), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s128>) = G_LOAD %0 :: (load (<2 x s128>), align 32, addrspace 1) %1:_(<2 x s128>) = G_LOAD %0 :: (load (<2 x s128>), align 32, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
... ...
--- ---
name: test_load_global_v2p1_align16 name: test_load_global_v2p1_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v2p1_align16 ; SI-LABEL: name: test_load_global_v2p1_align16
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; SI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; CI-HSA-LABEL: name: test_load_global_v2p1_align16 ; CI-HSA-LABEL: name: test_load_global_v2p1_align16
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; CI-MESA-LABEL: name: test_load_global_v2p1_align16 ; CI-MESA-LABEL: name: test_load_global_v2p1_align16
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; VI-LABEL: name: test_load_global_v2p1_align16 ; VI-LABEL: name: test_load_global_v2p1_align16
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX9-HSA-LABEL: name: test_load_global_v2p1_align16 ; GFX9-HSA-LABEL: name: test_load_global_v2p1_align16
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX9-MESA-LABEL: name: test_load_global_v2p1_align16 ; GFX9-MESA-LABEL: name: test_load_global_v2p1_align16
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 16, addrspace 1) %1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 16, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
name: test_load_global_v2p1_align8 name: test_load_global_v2p1_align8
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v2p1_align8 ; SI-LABEL: name: test_load_global_v2p1_align8
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 8, addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
; SI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; CI-HSA-LABEL: name: test_load_global_v2p1_align8 ; CI-HSA-LABEL: name: test_load_global_v2p1_align8
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 8, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; CI-MESA-LABEL: name: test_load_global_v2p1_align8 ; CI-MESA-LABEL: name: test_load_global_v2p1_align8
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 8, addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; VI-LABEL: name: test_load_global_v2p1_align8 ; VI-LABEL: name: test_load_global_v2p1_align8
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 8, addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX9-HSA-LABEL: name: test_load_global_v2p1_align8 ; GFX9-HSA-LABEL: name: test_load_global_v2p1_align8
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 8, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX9-MESA-LABEL: name: test_load_global_v2p1_align8 ; GFX9-MESA-LABEL: name: test_load_global_v2p1_align8
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 8, addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 8, addrspace 1) %1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 8, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
name: test_load_global_v2p1_align4 name: test_load_global_v2p1_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v2p1_align4 ; SI-LABEL: name: test_load_global_v2p1_align4
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
; SI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; CI-HSA-LABEL: name: test_load_global_v2p1_align4 ; CI-HSA-LABEL: name: test_load_global_v2p1_align4
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; CI-MESA-LABEL: name: test_load_global_v2p1_align4 ; CI-MESA-LABEL: name: test_load_global_v2p1_align4
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; VI-LABEL: name: test_load_global_v2p1_align4 ; VI-LABEL: name: test_load_global_v2p1_align4
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX9-HSA-LABEL: name: test_load_global_v2p1_align4 ; GFX9-HSA-LABEL: name: test_load_global_v2p1_align4
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX9-MESA-LABEL: name: test_load_global_v2p1_align4 ; GFX9-MESA-LABEL: name: test_load_global_v2p1_align4
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 4, addrspace 1) %1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 4, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
▲ Show 20 Lines • Show All 101 Lines • ▼ Show 20 Lines bb.0:
; SI: [[AND15:%[0-9]+]]:_(s32) = G_AND [[COPY16]], [[C3]] ; SI: [[AND15:%[0-9]+]]:_(s32) = G_AND [[COPY16]], [[C3]]
; SI: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C6]](s32) ; SI: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C6]](s32)
; SI: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]] ; SI: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]]
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
; SI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; CI-HSA-LABEL: name: test_load_global_v2p1_align1 ; CI-HSA-LABEL: name: test_load_global_v2p1_align1
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 1, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 1, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; CI-MESA-LABEL: name: test_load_global_v2p1_align1 ; CI-MESA-LABEL: name: test_load_global_v2p1_align1
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1 ; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; CI-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1) ; CI-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
▲ Show 20 Lines • Show All 192 Lines • ▼ Show 20 Lines bb.0:
; VI: [[AND15:%[0-9]+]]:_(s32) = G_AND [[COPY16]], [[C3]] ; VI: [[AND15:%[0-9]+]]:_(s32) = G_AND [[COPY16]], [[C3]]
; VI: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C6]](s32) ; VI: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C6]](s32)
; VI: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]] ; VI: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]]
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX9-HSA-LABEL: name: test_load_global_v2p1_align1 ; GFX9-HSA-LABEL: name: test_load_global_v2p1_align1
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 1, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 1, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX9-MESA-LABEL: name: test_load_global_v2p1_align1 ; GFX9-MESA-LABEL: name: test_load_global_v2p1_align1
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1 ; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1) ; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
▲ Show 20 Lines • Show All 101 Lines • ▼ Show 20 Lines
--- ---
name: test_load_global_v4p1_align8 name: test_load_global_v4p1_align8
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v4p1_align8 ; SI-LABEL: name: test_load_global_v4p1_align8
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x p1>), align 8, addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1)
; SI: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>) ; SI: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>)
; SI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>)
; CI-HSA-LABEL: name: test_load_global_v4p1_align8 ; CI-HSA-LABEL: name: test_load_global_v4p1_align8
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x p1>), align 8, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>)
; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>)
; CI-MESA-LABEL: name: test_load_global_v4p1_align8 ; CI-MESA-LABEL: name: test_load_global_v4p1_align8
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x p1>), align 8, addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>)
; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>)
; VI-LABEL: name: test_load_global_v4p1_align8 ; VI-LABEL: name: test_load_global_v4p1_align8
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x p1>), align 8, addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>)
; GFX9-HSA-LABEL: name: test_load_global_v4p1_align8 ; GFX9-HSA-LABEL: name: test_load_global_v4p1_align8
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x p1>), align 8, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>)
; GFX9-MESA-LABEL: name: test_load_global_v4p1_align8 ; GFX9-MESA-LABEL: name: test_load_global_v4p1_align8
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x p1>), align 8, addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x p1>) = G_LOAD %0 :: (load (<4 x p1>), align 8, addrspace 1) %1:_(<4 x p1>) = G_LOAD %0 :: (load (<4 x p1>), align 8, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
... ...
--- ---
▲ Show 20 Lines • Show All 1,349 Lines • ▼ Show 20 Lines bb.0:
; SI: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR11]](s32), [[OR14]](s32), [[OR17]](s32) ; SI: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR11]](s32), [[OR14]](s32), [[OR17]](s32)
; SI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>) ; SI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
; SI: [[COPY25:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; SI: [[COPY25:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; SI: [[COPY26:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; SI: [[COPY26:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; SI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY25]](s96) ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY25]](s96)
; SI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY26]](s96) ; SI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY26]](s96)
; CI-HSA-LABEL: name: test_global_v2s96_align1 ; CI-HSA-LABEL: name: test_global_v2s96_align1
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 1, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; CI-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; CI-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; CI-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; CI-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 1, addrspace 1) ; CI-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 1, addrspace 1)
; CI-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; CI-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; CI-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; CI-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; CI-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; CI-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; CI-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; CI-MESA-LABEL: name: test_global_v2s96_align1 ; CI-MESA-LABEL: name: test_global_v2s96_align1
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
▲ Show 20 Lines • Show All 293 Lines • ▼ Show 20 Lines bb.0:
; VI: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR11]](s32), [[OR14]](s32), [[OR17]](s32) ; VI: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR11]](s32), [[OR14]](s32), [[OR17]](s32)
; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>) ; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
; VI: [[COPY25:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; VI: [[COPY25:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; VI: [[COPY26:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; VI: [[COPY26:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY25]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY25]](s96)
; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY26]](s96) ; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY26]](s96)
; GFX9-HSA-LABEL: name: test_global_v2s96_align1 ; GFX9-HSA-LABEL: name: test_global_v2s96_align1
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 1, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; GFX9-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; GFX9-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; GFX9-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 1, addrspace 1) ; GFX9-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 1, addrspace 1)
; GFX9-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; GFX9-MESA-LABEL: name: test_global_v2s96_align1 ; GFX9-MESA-LABEL: name: test_global_v2s96_align1
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
▲ Show 20 Lines • Show All 231 Lines • ▼ Show 20 Lines bb.0:
; SI: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR3]](s32), [[OR4]](s32), [[OR5]](s32) ; SI: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR3]](s32), [[OR4]](s32), [[OR5]](s32)
; SI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>) ; SI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
; SI: [[COPY13:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; SI: [[COPY13:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; SI: [[COPY14:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; SI: [[COPY14:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; SI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY13]](s96) ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY13]](s96)
; SI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY14]](s96) ; SI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY14]](s96)
; CI-HSA-LABEL: name: test_global_v2s96_align2 ; CI-HSA-LABEL: name: test_global_v2s96_align2
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 2, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; CI-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; CI-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; CI-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; CI-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 2, addrspace 1) ; CI-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 2, addrspace 1)
; CI-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; CI-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; CI-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; CI-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; CI-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; CI-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; CI-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; CI-MESA-LABEL: name: test_global_v2s96_align2 ; CI-MESA-LABEL: name: test_global_v2s96_align2
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
▲ Show 20 Lines • Show All 141 Lines • ▼ Show 20 Lines bb.0:
; VI: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR3]](s32), [[OR4]](s32), [[OR5]](s32) ; VI: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR3]](s32), [[OR4]](s32), [[OR5]](s32)
; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>) ; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
; VI: [[COPY13:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; VI: [[COPY13:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; VI: [[COPY14:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; VI: [[COPY14:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY13]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY13]](s96)
; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY14]](s96) ; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY14]](s96)
; GFX9-HSA-LABEL: name: test_global_v2s96_align2 ; GFX9-HSA-LABEL: name: test_global_v2s96_align2
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 2, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; GFX9-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; GFX9-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; GFX9-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 2, addrspace 1) ; GFX9-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 2, addrspace 1)
; GFX9-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; GFX9-MESA-LABEL: name: test_global_v2s96_align2 ; GFX9-MESA-LABEL: name: test_global_v2s96_align2
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
▲ Show 20 Lines • Show All 101 Lines • ▼ Show 20 Lines bb.0:
; SI: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[UV3]](s32), [[LOAD3]](s32) ; SI: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[UV3]](s32), [[LOAD3]](s32)
; SI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>) ; SI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
; SI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; SI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; SI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; SI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; SI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; SI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; SI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; CI-HSA-LABEL: name: test_global_v2s96_align4 ; CI-HSA-LABEL: name: test_global_v2s96_align4
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; CI-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; CI-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; CI-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; CI-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) ; CI-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
; CI-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; CI-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; CI-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; CI-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; CI-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; CI-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; CI-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; CI-MESA-LABEL: name: test_global_v2s96_align4 ; CI-MESA-LABEL: name: test_global_v2s96_align4
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; CI-MESA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) ; CI-MESA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
; CI-MESA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; CI-MESA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; CI-MESA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI-MESA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; CI-MESA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; CI-MESA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; CI-MESA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; CI-MESA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; VI-LABEL: name: test_global_v2s96_align4 ; VI-LABEL: name: test_global_v2s96_align4
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; VI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; VI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; VI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; VI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; VI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) ; VI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; VI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; VI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; GFX9-HSA-LABEL: name: test_global_v2s96_align4 ; GFX9-HSA-LABEL: name: test_global_v2s96_align4
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; GFX9-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; GFX9-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; GFX9-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) ; GFX9-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
; GFX9-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; GFX9-MESA-LABEL: name: test_global_v2s96_align4 ; GFX9-MESA-LABEL: name: test_global_v2s96_align4
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) ; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
; GFX9-MESA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-MESA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; GFX9-MESA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-MESA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; GFX9-MESA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; GFX9-MESA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-MESA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-MESA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 4, addrspace 1) %1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 4, addrspace 1)
%2:_(s96) = G_EXTRACT %1, 0 %2:_(s96) = G_EXTRACT %1, 0
Show All 23 Lines bb.0:
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD2]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD2]](s32)
; SI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; SI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; SI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; SI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; SI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; SI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; SI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; SI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; SI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; CI-HSA-LABEL: name: test_global_v2s96_align16 ; CI-HSA-LABEL: name: test_global_v2s96_align16
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; CI-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; CI-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; CI-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; CI-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) ; CI-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
; CI-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; CI-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; CI-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; CI-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; CI-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; CI-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; CI-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; CI-MESA-LABEL: name: test_global_v2s96_align16 ; CI-MESA-LABEL: name: test_global_v2s96_align16
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; CI-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; CI-MESA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) ; CI-MESA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
; CI-MESA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; CI-MESA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; CI-MESA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI-MESA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; CI-MESA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; CI-MESA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; CI-MESA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; CI-MESA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; VI-LABEL: name: test_global_v2s96_align16 ; VI-LABEL: name: test_global_v2s96_align16
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; VI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; VI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; VI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; VI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; VI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) ; VI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; VI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; VI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; GFX9-HSA-LABEL: name: test_global_v2s96_align16 ; GFX9-HSA-LABEL: name: test_global_v2s96_align16
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; GFX9-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; GFX9-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; GFX9-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) ; GFX9-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
; GFX9-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; GFX9-MESA-LABEL: name: test_global_v2s96_align16 ; GFX9-MESA-LABEL: name: test_global_v2s96_align16
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) ; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
; GFX9-MESA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-MESA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; GFX9-MESA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-MESA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; GFX9-MESA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; GFX9-MESA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-MESA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-MESA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 16, addrspace 1) %1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 16, addrspace 1)
%2:_(s96) = G_EXTRACT %1, 0 %2:_(s96) = G_EXTRACT %1, 0
%3:_(s96) = G_EXTRACT %1, 96 %3:_(s96) = G_EXTRACT %1, 96
$vgpr0_vgpr1_vgpr2 = COPY %2 $vgpr0_vgpr1_vgpr2 = COPY %2
$vgpr3_vgpr4_vgpr5 = COPY %3 $vgpr3_vgpr4_vgpr5 = COPY %3
... ...
--- ---
name: test_load_global_v32s1_align4 name: test_load_global_v32s1_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v32s1_align4 ; SI-LABEL: name: test_load_global_v32s1_align4
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<32 x s1>), addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2 ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3 ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; SI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; SI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
▲ Show 20 Lines • Show All 83 Lines • ▼ Show 20 Lines bb.0:
; SI: [[COPY30:%[0-9]+]]:_(s32) = COPY [[LSHR28]](s32) ; SI: [[COPY30:%[0-9]+]]:_(s32) = COPY [[LSHR28]](s32)
; SI: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR29]](s32) ; SI: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR29]](s32)
; SI: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR30]](s32) ; SI: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR30]](s32)
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<32 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32), [[COPY12]](s32), [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32), [[COPY17]](s32), [[COPY18]](s32), [[COPY19]](s32), [[COPY20]](s32), [[COPY21]](s32), [[COPY22]](s32), [[COPY23]](s32), [[COPY24]](s32), [[COPY25]](s32), [[COPY26]](s32), [[COPY27]](s32), [[COPY28]](s32), [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<32 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32), [[COPY12]](s32), [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32), [[COPY17]](s32), [[COPY18]](s32), [[COPY19]](s32), [[COPY20]](s32), [[COPY21]](s32), [[COPY22]](s32), [[COPY23]](s32), [[COPY24]](s32), [[COPY25]](s32), [[COPY26]](s32), [[COPY27]](s32), [[COPY28]](s32), [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32)
; SI: [[TRUNC:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[BUILD_VECTOR]](<32 x s32>) ; SI: [[TRUNC:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[BUILD_VECTOR]](<32 x s32>)
; SI: $vgpr0 = COPY [[TRUNC]](<32 x s1>) ; SI: $vgpr0 = COPY [[TRUNC]](<32 x s1>)
; CI-HSA-LABEL: name: test_load_global_v32s1_align4 ; CI-HSA-LABEL: name: test_load_global_v32s1_align4
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<32 x s1>), addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2 ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3 ; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI-HSA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; CI-HSA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
; CI-HSA: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32) ; CI-HSA: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
▲ Show 20 Lines • Show All 83 Lines • ▼ Show 20 Lines bb.0:
; CI-HSA: [[COPY30:%[0-9]+]]:_(s32) = COPY [[LSHR28]](s32) ; CI-HSA: [[COPY30:%[0-9]+]]:_(s32) = COPY [[LSHR28]](s32)
; CI-HSA: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR29]](s32) ; CI-HSA: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR29]](s32)
; CI-HSA: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR30]](s32) ; CI-HSA: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR30]](s32)
; CI-HSA: [[BUILD_VECTOR:%[0-9]+]]:_(<32 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32), [[COPY12]](s32), [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32), [[COPY17]](s32), [[COPY18]](s32), [[COPY19]](s32), [[COPY20]](s32), [[COPY21]](s32), [[COPY22]](s32), [[COPY23]](s32), [[COPY24]](s32), [[COPY25]](s32), [[COPY26]](s32), [[COPY27]](s32), [[COPY28]](s32), [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32) ; CI-HSA: [[BUILD_VECTOR:%[0-9]+]]:_(<32 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32), [[COPY12]](s32), [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32), [[COPY17]](s32), [[COPY18]](s32), [[COPY19]](s32), [[COPY20]](s32), [[COPY21]](s32), [[COPY22]](s32), [[COPY23]](s32), [[COPY24]](s32), [[COPY25]](s32), [[COPY26]](s32), [[COPY27]](s32), [[COPY28]](s32), [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32)
; CI-HSA: [[TRUNC:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[BUILD_VECTOR]](<32 x s32>) ; CI-HSA: [[TRUNC:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[BUILD_VECTOR]](<32 x s32>)
; CI-HSA: $vgpr0 = COPY [[TRUNC]](<32 x s1>) ; CI-HSA: $vgpr0 = COPY [[TRUNC]](<32 x s1>)
; CI-MESA-LABEL: name: test_load_global_v32s1_align4 ; CI-MESA-LABEL: name: test_load_global_v32s1_align4
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<32 x s1>), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2 ; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
; CI-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3 ; CI-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
; CI-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI-MESA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; CI-MESA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
; CI-MESA: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32) ; CI-MESA: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
▲ Show 20 Lines • Show All 83 Lines • ▼ Show 20 Lines bb.0:
; CI-MESA: [[COPY30:%[0-9]+]]:_(s32) = COPY [[LSHR28]](s32) ; CI-MESA: [[COPY30:%[0-9]+]]:_(s32) = COPY [[LSHR28]](s32)
; CI-MESA: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR29]](s32) ; CI-MESA: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR29]](s32)
; CI-MESA: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR30]](s32) ; CI-MESA: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR30]](s32)
; CI-MESA: [[BUILD_VECTOR:%[0-9]+]]:_(<32 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32), [[COPY12]](s32), [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32), [[COPY17]](s32), [[COPY18]](s32), [[COPY19]](s32), [[COPY20]](s32), [[COPY21]](s32), [[COPY22]](s32), [[COPY23]](s32), [[COPY24]](s32), [[COPY25]](s32), [[COPY26]](s32), [[COPY27]](s32), [[COPY28]](s32), [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32) ; CI-MESA: [[BUILD_VECTOR:%[0-9]+]]:_(<32 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32), [[COPY12]](s32), [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32), [[COPY17]](s32), [[COPY18]](s32), [[COPY19]](s32), [[COPY20]](s32), [[COPY21]](s32), [[COPY22]](s32), [[COPY23]](s32), [[COPY24]](s32), [[COPY25]](s32), [[COPY26]](s32), [[COPY27]](s32), [[COPY28]](s32), [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32)
; CI-MESA: [[TRUNC:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[BUILD_VECTOR]](<32 x s32>) ; CI-MESA: [[TRUNC:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[BUILD_VECTOR]](<32 x s32>)
; CI-MESA: $vgpr0 = COPY [[TRUNC]](<32 x s1>) ; CI-MESA: $vgpr0 = COPY [[TRUNC]](<32 x s1>)
; VI-LABEL: name: test_load_global_v32s1_align4 ; VI-LABEL: name: test_load_global_v32s1_align4
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<32 x s1>), addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; VI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; VI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
▲ Show 20 Lines • Show All 83 Lines • ▼ Show 20 Lines bb.0:
; VI: [[COPY30:%[0-9]+]]:_(s32) = COPY [[LSHR28]](s32) ; VI: [[COPY30:%[0-9]+]]:_(s32) = COPY [[LSHR28]](s32)
; VI: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR29]](s32) ; VI: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR29]](s32)
; VI: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR30]](s32) ; VI: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR30]](s32)
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<32 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32), [[COPY12]](s32), [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32), [[COPY17]](s32), [[COPY18]](s32), [[COPY19]](s32), [[COPY20]](s32), [[COPY21]](s32), [[COPY22]](s32), [[COPY23]](s32), [[COPY24]](s32), [[COPY25]](s32), [[COPY26]](s32), [[COPY27]](s32), [[COPY28]](s32), [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<32 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32), [[COPY12]](s32), [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32), [[COPY16]](s32), [[COPY17]](s32), [[COPY18]](s32), [[COPY19]](s32), [[COPY20]](s32), [[COPY21]](s32), [[COPY22]](s32), [[COPY23]](s32), [[COPY24]](s32), [[COPY25]](s32), [[COPY26]](s32), [[COPY27]](s32), [[COPY28]](s32), [[COPY29]](s32), [[COPY30]](s32), [[COPY31]](s32), [[COPY32]](s32)
; VI: [[TRUNC:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[BUILD_VECTOR]](<32 x s32>) ; VI: [[TRUNC:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[BUILD_VECTOR]](<32 x s32>)
; VI: $vgpr0 = COPY [[TRUNC]](<32 x s1>) ; VI: $vgpr0 = COPY [[TRUNC]](<32 x s1>)
; GFX9-HSA-LABEL: name: test_load_global_v32s1_align4 ; GFX9-HSA-LABEL: name: test_load_global_v32s1_align4
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<32 x s1>), addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2 ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3 ; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9-HSA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; GFX9-HSA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
; GFX9-HSA: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32) ; GFX9-HSA: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
▲ Show 20 Lines • Show All 99 Lines • ▼ Show 20 Lines bb.0:
; GFX9-HSA: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR29]](s32) ; GFX9-HSA: [[COPY31:%[0-9]+]]:_(s32) = COPY [[LSHR29]](s32)
; GFX9-HSA: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR30]](s32) ; GFX9-HSA: [[COPY32:%[0-9]+]]:_(s32) = COPY [[LSHR30]](s32)
; GFX9-HSA: [[BUILD_VECTOR_TRUNC15:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY31]](s32), [[COPY32]](s32) ; GFX9-HSA: [[BUILD_VECTOR_TRUNC15:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY31]](s32), [[COPY32]](s32)
; GFX9-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>), [[BUILD_VECTOR_TRUNC4]](<2 x s16>), [[BUILD_VECTOR_TRUNC5]](<2 x s16>), [[BUILD_VECTOR_TRUNC6]](<2 x s16>), [[BUILD_VECTOR_TRUNC7]](<2 x s16>), [[BUILD_VECTOR_TRUNC8]](<2 x s16>), [[BUILD_VECTOR_TRUNC9]](<2 x s16>), [[BUILD_VECTOR_TRUNC10]](<2 x s16>), [[BUILD_VECTOR_TRUNC11]](<2 x s16>), [[BUILD_VECTOR_TRUNC12]](<2 x s16>), [[BUILD_VECTOR_TRUNC13]](<2 x s16>), [[BUILD_VECTOR_TRUNC14]](<2 x s16>), [[BUILD_VECTOR_TRUNC15]](<2 x s16>) ; GFX9-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>), [[BUILD_VECTOR_TRUNC4]](<2 x s16>), [[BUILD_VECTOR_TRUNC5]](<2 x s16>), [[BUILD_VECTOR_TRUNC6]](<2 x s16>), [[BUILD_VECTOR_TRUNC7]](<2 x s16>), [[BUILD_VECTOR_TRUNC8]](<2 x s16>), [[BUILD_VECTOR_TRUNC9]](<2 x s16>), [[BUILD_VECTOR_TRUNC10]](<2 x s16>), [[BUILD_VECTOR_TRUNC11]](<2 x s16>), [[BUILD_VECTOR_TRUNC12]](<2 x s16>), [[BUILD_VECTOR_TRUNC13]](<2 x s16>), [[BUILD_VECTOR_TRUNC14]](<2 x s16>), [[BUILD_VECTOR_TRUNC15]](<2 x s16>)
; GFX9-HSA: [[TRUNC:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[CONCAT_VECTORS]](<32 x s16>) ; GFX9-HSA: [[TRUNC:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[CONCAT_VECTORS]](<32 x s16>)
; GFX9-HSA: $vgpr0 = COPY [[TRUNC]](<32 x s1>) ; GFX9-HSA: $vgpr0 = COPY [[TRUNC]](<32 x s1>)
; GFX9-MESA-LABEL: name: test_load_global_v32s1_align4 ; GFX9-MESA-LABEL: name: test_load_global_v32s1_align4
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<32 x s1>), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2 ; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
; GFX9-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3 ; GFX9-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
; GFX9-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9-MESA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; GFX9-MESA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
; GFX9-MESA: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32) ; GFX9-MESA: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
▲ Show 20 Lines • Show All 110 Lines • ▼ Show 20 Lines
--- ---
name: test_load_global_v8s4_align4 name: test_load_global_v8s4_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1 liveins: $vgpr0_vgpr1
; SI-LABEL: name: test_load_global_v8s4_align4 ; SI-LABEL: name: test_load_global_v8s4_align4
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<8 x s4>), addrspace 1) ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; SI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
Show All 11 Lines bb.0:
; SI: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; SI: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; SI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; SI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; SI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) ; SI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32)
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32)
; SI: [[TRUNC:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[BUILD_VECTOR]](<8 x s32>) ; SI: [[TRUNC:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[BUILD_VECTOR]](<8 x s32>)
; SI: $vgpr0 = COPY [[TRUNC]](<8 x s4>) ; SI: $vgpr0 = COPY [[TRUNC]](<8 x s4>)
; CI-HSA-LABEL: name: test_load_global_v8s4_align4 ; CI-HSA-LABEL: name: test_load_global_v8s4_align4
; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<8 x s4>), addrspace 1) ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI-HSA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-HSA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-HSA: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32) ; CI-HSA: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
Show All 11 Lines bb.0:
; CI-HSA: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; CI-HSA: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; CI-HSA: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; CI-HSA: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; CI-HSA: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) ; CI-HSA: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32)
; CI-HSA: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32) ; CI-HSA: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32)
; CI-HSA: [[TRUNC:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[BUILD_VECTOR]](<8 x s32>) ; CI-HSA: [[TRUNC:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[BUILD_VECTOR]](<8 x s32>)
; CI-HSA: $vgpr0 = COPY [[TRUNC]](<8 x s4>) ; CI-HSA: $vgpr0 = COPY [[TRUNC]](<8 x s4>)
; CI-MESA-LABEL: name: test_load_global_v8s4_align4 ; CI-MESA-LABEL: name: test_load_global_v8s4_align4
; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<8 x s4>), addrspace 1) ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; CI-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
; CI-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI-MESA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-MESA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-MESA: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32) ; CI-MESA: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
Show All 11 Lines bb.0:
; CI-MESA: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; CI-MESA: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; CI-MESA: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; CI-MESA: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; CI-MESA: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) ; CI-MESA: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32)
; CI-MESA: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32) ; CI-MESA: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32)
; CI-MESA: [[TRUNC:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[BUILD_VECTOR]](<8 x s32>) ; CI-MESA: [[TRUNC:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[BUILD_VECTOR]](<8 x s32>)
; CI-MESA: $vgpr0 = COPY [[TRUNC]](<8 x s4>) ; CI-MESA: $vgpr0 = COPY [[TRUNC]](<8 x s4>)
; VI-LABEL: name: test_load_global_v8s4_align4 ; VI-LABEL: name: test_load_global_v8s4_align4
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<8 x s4>), addrspace 1) ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; VI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
Show All 11 Lines bb.0:
; VI: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; VI: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; VI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) ; VI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32)
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32)
; VI: [[TRUNC:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[BUILD_VECTOR]](<8 x s32>) ; VI: [[TRUNC:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[BUILD_VECTOR]](<8 x s32>)
; VI: $vgpr0 = COPY [[TRUNC]](<8 x s4>) ; VI: $vgpr0 = COPY [[TRUNC]](<8 x s4>)
; GFX9-HSA-LABEL: name: test_load_global_v8s4_align4 ; GFX9-HSA-LABEL: name: test_load_global_v8s4_align4
; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<8 x s4>), addrspace 1) ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9-HSA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-HSA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-HSA: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32) ; GFX9-HSA: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
Show All 15 Lines bb.0:
; GFX9-HSA: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; GFX9-HSA: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; GFX9-HSA: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) ; GFX9-HSA: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32)
; GFX9-HSA: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY7]](s32), [[COPY8]](s32) ; GFX9-HSA: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY7]](s32), [[COPY8]](s32)
; GFX9-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>) ; GFX9-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>)
; GFX9-HSA: [[TRUNC:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[CONCAT_VECTORS]](<8 x s16>) ; GFX9-HSA: [[TRUNC:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[CONCAT_VECTORS]](<8 x s16>)
; GFX9-HSA: $vgpr0 = COPY [[TRUNC]](<8 x s4>) ; GFX9-HSA: $vgpr0 = COPY [[TRUNC]](<8 x s4>)
; GFX9-MESA-LABEL: name: test_load_global_v8s4_align4 ; GFX9-MESA-LABEL: name: test_load_global_v8s4_align4
; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<8 x s4>), addrspace 1) ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9-MESA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX9-MESA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
; GFX9-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9-MESA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9-MESA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-MESA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-MESA: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32) ; GFX9-MESA: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
Show All 25 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-local.mir

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 2,711 Lines • ▼ Show 20 Lines bb.0:
; GFX9: [[AND11:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C3]] ; GFX9: [[AND11:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C3]]
; GFX9: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C6]](s32) ; GFX9: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C6]](s32)
; GFX9: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]] ; GFX9: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]]
; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32) ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-UNALIGNED-LABEL: name: test_load_local_s96_align16 ; GFX9-UNALIGNED-LABEL: name: test_load_local_s96_align16
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 1, addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 1, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX10-LABEL: name: test_load_local_s96_align16 ; GFX10-LABEL: name: test_load_local_s96_align16
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3)
; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3) ; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
▲ Show 20 Lines • Show All 205 Lines • ▼ Show 20 Lines bb.0:
; GFX9: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX9: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX9: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 8, align 8, addrspace 3) ; GFX9: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 8, align 8, addrspace 3)
; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32) ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32)
; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-UNALIGNED-LABEL: name: test_load_local_s96_align8 ; GFX9-UNALIGNED-LABEL: name: test_load_local_s96_align8
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 8, addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 8, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX10-LABEL: name: test_load_local_s96_align8 ; GFX10-LABEL: name: test_load_local_s96_align8
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 8, align 8, addrspace 3) ; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 8, align 8, addrspace 3)
▲ Show 20 Lines • Show All 69 Lines • ▼ Show 20 Lines bb.0:
; GFX9: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX9: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX9: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 8, addrspace 3) ; GFX9: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 8, addrspace 3)
; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32) ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32)
; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-UNALIGNED-LABEL: name: test_load_local_s96_align4 ; GFX9-UNALIGNED-LABEL: name: test_load_local_s96_align4
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 4, addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 4, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX10-LABEL: name: test_load_local_s96_align4 ; GFX10-LABEL: name: test_load_local_s96_align4
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), align 4, addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), align 4, addrspace 3)
; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 8, addrspace 3) ; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 8, addrspace 3)
▲ Show 20 Lines • Show All 214 Lines • ▼ Show 20 Lines bb.0:
; GFX9: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C1]] ; GFX9: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C1]]
; GFX9: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C2]](s32) ; GFX9: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C2]](s32)
; GFX9: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]] ; GFX9: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32) ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-UNALIGNED-LABEL: name: test_load_local_s96_align2 ; GFX9-UNALIGNED-LABEL: name: test_load_local_s96_align2
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 2, addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 2, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX10-LABEL: name: test_load_local_s96_align2 ; GFX10-LABEL: name: test_load_local_s96_align2
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3) ; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
▲ Show 20 Lines • Show All 467 Lines • ▼ Show 20 Lines bb.0:
; GFX9: [[AND11:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C3]] ; GFX9: [[AND11:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C3]]
; GFX9: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C6]](s32) ; GFX9: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C6]](s32)
; GFX9: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]] ; GFX9: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]]
; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32) ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX9-UNALIGNED-LABEL: name: test_load_local_s96_align1 ; GFX9-UNALIGNED-LABEL: name: test_load_local_s96_align1
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 1, addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 1, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
; GFX10-LABEL: name: test_load_local_s96_align1 ; GFX10-LABEL: name: test_load_local_s96_align1
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3)
; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3) ; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
▲ Show 20 Lines • Show All 662 Lines • ▼ Show 20 Lines bb.0:
; GFX9: [[AND15:%[0-9]+]]:_(s32) = G_AND [[COPY16]], [[C3]] ; GFX9: [[AND15:%[0-9]+]]:_(s32) = G_AND [[COPY16]], [[C3]]
; GFX9: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C6]](s32) ; GFX9: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C6]](s32)
; GFX9: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]] ; GFX9: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]]
; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32) ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX9-UNALIGNED-LABEL: name: test_load_local_s128_align16 ; GFX9-UNALIGNED-LABEL: name: test_load_local_s128_align16
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 1, addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 1, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX10-LABEL: name: test_load_local_s128_align16 ; GFX10-LABEL: name: test_load_local_s128_align16
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3)
; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3) ; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
▲ Show 20 Lines • Show All 219 Lines • ▼ Show 20 Lines bb.0:
; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; CI: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; CI: [[LOAD1:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<2 x s32>) from unknown-address + 8, addrspace 3) ; CI: [[LOAD1:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<2 x s32>) from unknown-address + 8, addrspace 3)
; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s32>) = G_CONCAT_VECTORS [[LOAD]](<2 x s32>), [[LOAD1]](<2 x s32>) ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s32>) = G_CONCAT_VECTORS [[LOAD]](<2 x s32>), [[LOAD1]](<2 x s32>)
; CI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[CONCAT_VECTORS]](<4 x s32>) ; CI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[CONCAT_VECTORS]](<4 x s32>)
; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; CI-DS128-LABEL: name: test_load_local_s128_align8 ; CI-DS128-LABEL: name: test_load_local_s128_align8
; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; CI-DS128: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 8, addrspace 3) ; CI-DS128: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3)
; CI-DS128: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-DS128: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; CI-DS128: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI-DS128: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; VI-LABEL: name: test_load_local_s128_align8 ; VI-LABEL: name: test_load_local_s128_align8
; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 8, addrspace 3) ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3)
; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX9-LABEL: name: test_load_local_s128_align8 ; GFX9-LABEL: name: test_load_local_s128_align8
; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 8, addrspace 3) ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3)
; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX9-UNALIGNED-LABEL: name: test_load_local_s128_align8 ; GFX9-UNALIGNED-LABEL: name: test_load_local_s128_align8
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 8, addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX10-LABEL: name: test_load_local_s128_align8 ; GFX10-LABEL: name: test_load_local_s128_align8
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 8, addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3)
; GFX10: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX10: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX10: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX10: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX10-UNALIGNED-LABEL: name: test_load_local_s128_align8 ; GFX10-UNALIGNED-LABEL: name: test_load_local_s128_align8
; GFX10-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 8, addrspace 3) ; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3)
; GFX10-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX10-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX10-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX10-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p3) = COPY $vgpr0 %0:_(p3) = COPY $vgpr0
%1:_(s128) = G_LOAD %0 :: (load (s128), align 8, addrspace 3) %1:_(s128) = G_LOAD %0 :: (load (s128), align 8, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
... ...
--- ---
▲ Show 20 Lines • Show All 62 Lines • ▼ Show 20 Lines bb.0:
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
; GFX9: [[PTR_ADD2:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C2]](s32) ; GFX9: [[PTR_ADD2:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C2]](s32)
; GFX9: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s32) from unknown-address + 12, addrspace 3) ; GFX9: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s32) from unknown-address + 12, addrspace 3)
; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LOAD1]](s32), [[LOAD2]](s32), [[LOAD3]](s32) ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LOAD1]](s32), [[LOAD2]](s32), [[LOAD3]](s32)
; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX9-UNALIGNED-LABEL: name: test_load_local_s128_align4 ; GFX9-UNALIGNED-LABEL: name: test_load_local_s128_align4
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 4, addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 4, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX10-LABEL: name: test_load_local_s128_align4 ; GFX10-LABEL: name: test_load_local_s128_align4
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 4, addrspace 3) ; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 4, addrspace 3)
▲ Show 20 Lines • Show All 284 Lines • ▼ Show 20 Lines bb.0:
; GFX9: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C2]](s32) ; GFX9: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C2]](s32)
; GFX9: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]] ; GFX9: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
; GFX9: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR3]](s32) ; GFX9: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR3]](s32)
; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s32>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s32>), [[BUILD_VECTOR1]](<2 x s32>) ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s32>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s32>), [[BUILD_VECTOR1]](<2 x s32>)
; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[CONCAT_VECTORS]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[CONCAT_VECTORS]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX9-UNALIGNED-LABEL: name: test_load_local_s128_align2 ; GFX9-UNALIGNED-LABEL: name: test_load_local_s128_align2
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 2, addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 2, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX10-LABEL: name: test_load_local_s128_align2 ; GFX10-LABEL: name: test_load_local_s128_align2
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3) ; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
▲ Show 20 Lines • Show All 608 Lines • ▼ Show 20 Lines bb.0:
; GFX9: [[AND15:%[0-9]+]]:_(s32) = G_AND [[COPY16]], [[C3]] ; GFX9: [[AND15:%[0-9]+]]:_(s32) = G_AND [[COPY16]], [[C3]]
; GFX9: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C6]](s32) ; GFX9: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C6]](s32)
; GFX9: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]] ; GFX9: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]]
; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32) ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX9-UNALIGNED-LABEL: name: test_load_local_s128_align1 ; GFX9-UNALIGNED-LABEL: name: test_load_local_s128_align1
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 1, addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 1, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
; GFX10-LABEL: name: test_load_local_s128_align1 ; GFX10-LABEL: name: test_load_local_s128_align1
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3)
; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3) ; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
▲ Show 20 Lines • Show All 1,782 Lines • ▼ Show 20 Lines
--- ---
name: test_load_local_v2s8_align2 name: test_load_local_v2s8_align2
body: | body: |
bb.0: bb.0:
liveins: $vgpr0 liveins: $vgpr0
; SI-LABEL: name: test_load_local_v2s8_align2 ; SI-LABEL: name: test_load_local_v2s8_align2
; SI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; SI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), addrspace 3) ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; SI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; SI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; SI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; SI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32) ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
; SI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; SI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]] ; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]]
; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32) ; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; SI: $vgpr0 = COPY [[ANYEXT]](s32) ; SI: $vgpr0 = COPY [[ANYEXT]](s32)
; CI-LABEL: name: test_load_local_v2s8_align2 ; CI-LABEL: name: test_load_local_v2s8_align2
; CI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; CI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), addrspace 3) ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; CI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; CI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; CI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; CI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; CI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32) ; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
; CI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; CI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]] ; CI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]]
; CI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32) ; CI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
; CI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; CI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; CI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; CI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; CI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; CI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; CI: $vgpr0 = COPY [[ANYEXT]](s32) ; CI: $vgpr0 = COPY [[ANYEXT]](s32)
; CI-DS128-LABEL: name: test_load_local_v2s8_align2 ; CI-DS128-LABEL: name: test_load_local_v2s8_align2
; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; CI-DS128: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), addrspace 3) ; CI-DS128: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
; CI-DS128: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-DS128: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-DS128: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-DS128: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI-DS128: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-DS128: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-DS128: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI-DS128: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI-DS128: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-DS128: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-DS128: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI-DS128: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI-DS128: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; CI-DS128: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; CI-DS128: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; CI-DS128: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; CI-DS128: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; CI-DS128: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; CI-DS128: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32) ; CI-DS128: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
; CI-DS128: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; CI-DS128: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; CI-DS128: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI-DS128: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI-DS128: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]] ; CI-DS128: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]]
; CI-DS128: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32) ; CI-DS128: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
; CI-DS128: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; CI-DS128: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; CI-DS128: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; CI-DS128: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; CI-DS128: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; CI-DS128: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; CI-DS128: $vgpr0 = COPY [[ANYEXT]](s32) ; CI-DS128: $vgpr0 = COPY [[ANYEXT]](s32)
; VI-LABEL: name: test_load_local_v2s8_align2 ; VI-LABEL: name: test_load_local_v2s8_align2
; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), addrspace 3) ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; VI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; VI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]] ; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
; VI: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 ; VI: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16) ; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; VI: $vgpr0 = COPY [[ANYEXT]](s32)
; GFX9-LABEL: name: test_load_local_v2s8_align2 ; GFX9-LABEL: name: test_load_local_v2s8_align2
; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), addrspace 3) ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; GFX9: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; GFX9: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; GFX9: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX9: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]] ; GFX9: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
; GFX9: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 ; GFX9: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; GFX9: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16) ; GFX9: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
; GFX9: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; GFX9: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; GFX9: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; GFX9: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; GFX9: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9: $vgpr0 = COPY [[ANYEXT]](s32)
; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s8_align2 ; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s8_align2
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9-UNALIGNED: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; GFX9-UNALIGNED: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; GFX9-UNALIGNED: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; GFX9-UNALIGNED: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; GFX9-UNALIGNED: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; GFX9-UNALIGNED: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; GFX9-UNALIGNED: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX9-UNALIGNED: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX9-UNALIGNED: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]] ; GFX9-UNALIGNED: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
; GFX9-UNALIGNED: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 ; GFX9-UNALIGNED: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; GFX9-UNALIGNED: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16) ; GFX9-UNALIGNED: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
; GFX9-UNALIGNED: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; GFX9-UNALIGNED: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; GFX9-UNALIGNED: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; GFX9-UNALIGNED: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; GFX9-UNALIGNED: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-UNALIGNED: $vgpr0 = COPY [[ANYEXT]](s32)
; GFX10-LABEL: name: test_load_local_v2s8_align2 ; GFX10-LABEL: name: test_load_local_v2s8_align2
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX10: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX10: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX10: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX10: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX10: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX10: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX10: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX10: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX10: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX10: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; GFX10: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; GFX10: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; GFX10: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; GFX10: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; GFX10: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; GFX10: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]] ; GFX10: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
; GFX10: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 ; GFX10: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; GFX10: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16) ; GFX10: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
; GFX10: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; GFX10: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; GFX10: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; GFX10: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; GFX10: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX10: $vgpr0 = COPY [[ANYEXT]](s32)
; GFX10-UNALIGNED-LABEL: name: test_load_local_v2s8_align2 ; GFX10-UNALIGNED-LABEL: name: test_load_local_v2s8_align2
; GFX10-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), addrspace 3) ; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
; GFX10-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX10-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX10-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX10-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX10-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX10-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX10-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX10-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX10-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX10-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX10-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX10-UNALIGNED: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; GFX10-UNALIGNED: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; GFX10-UNALIGNED: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; GFX10-UNALIGNED: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
▲ Show 20 Lines • Show All 73 Lines • ▼ Show 20 Lines bb.0:
; GFX9: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX9: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX9: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3) ; GFX9: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LOAD1]](s32) ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LOAD1]](s32)
; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32) ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32)
; GFX9: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>) ; GFX9: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s8_align1 ; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s8_align1
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), align 1, addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
▲ Show 20 Lines • Show All 300 Lines • ▼ Show 20 Lines bb.0:
; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>) ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
; GFX9: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[UV]](<3 x s16>) ; GFX9: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[UV]](<3 x s16>)
; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
; GFX9: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF2]](<4 x s32>) ; GFX9: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF2]](<4 x s32>)
; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC1]], [[TRUNC]](<3 x s8>), 0 ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC1]], [[TRUNC]](<3 x s8>), 0
; GFX9: $vgpr0 = COPY [[INSERT]](<4 x s8>) ; GFX9: $vgpr0 = COPY [[INSERT]](<4 x s8>)
; GFX9-UNALIGNED-LABEL: name: test_load_local_v3s8_align1 ; GFX9-UNALIGNED-LABEL: name: test_load_local_v3s8_align1
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(s24) = G_LOAD [[COPY]](p3) :: (load (<3 x s8>), align 1, addrspace 3) ; GFX9-UNALIGNED: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(<3 x s8>) = G_BITCAST [[LOAD]](s24) ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
; GFX9-UNALIGNED: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF ; GFX9-UNALIGNED: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX9-UNALIGNED: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF]](<4 x s32>) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
; GFX9-UNALIGNED: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC]], [[BITCAST]](<3 x s8>), 0 ; GFX9-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-UNALIGNED: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
; GFX9-UNALIGNED: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
; GFX9-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
; GFX9-UNALIGNED: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C3]](s32)
; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s32) = COPY [[OR]](s32)
; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; GFX9-UNALIGNED: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY1]](s32), [[COPY2]](s32)
; GFX9-UNALIGNED: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; GFX9-UNALIGNED: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
; GFX9-UNALIGNED: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[DEF]](s32)
; GFX9-UNALIGNED: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
; GFX9-UNALIGNED: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[DEF1]](<2 x s16>)
; GFX9-UNALIGNED: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
; GFX9-UNALIGNED: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[UV]](<3 x s16>)
; GFX9-UNALIGNED: [[DEF2:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
; GFX9-UNALIGNED: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF2]](<4 x s32>)
; GFX9-UNALIGNED: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC1]], [[TRUNC]](<3 x s8>), 0
; GFX9-UNALIGNED: $vgpr0 = COPY [[INSERT]](<4 x s8>) ; GFX9-UNALIGNED: $vgpr0 = COPY [[INSERT]](<4 x s8>)
; GFX10-LABEL: name: test_load_local_v3s8_align1 ; GFX10-LABEL: name: test_load_local_v3s8_align1
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3)
; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3) ; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
; GFX10: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2 ; GFX10: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
▲ Show 20 Lines • Show All 46 Lines • ▼ Show 20 Lines
--- ---
name: test_load_local_v4s8_align4 name: test_load_local_v4s8_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0 liveins: $vgpr0
; SI-LABEL: name: test_load_local_v4s8_align4 ; SI-LABEL: name: test_load_local_v4s8_align4
; SI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; SI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<4 x s8>), addrspace 3) ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; SI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; SI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; SI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; SI: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; CI-LABEL: name: test_load_local_v4s8_align4 ; CI-LABEL: name: test_load_local_v4s8_align4
; CI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; CI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<4 x s8>), addrspace 3) ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; CI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; CI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; CI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; CI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; CI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; CI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; CI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; CI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; CI: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; CI-DS128-LABEL: name: test_load_local_v4s8_align4 ; CI-DS128-LABEL: name: test_load_local_v4s8_align4
; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; CI-DS128: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<4 x s8>), addrspace 3) ; CI-DS128: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
; CI-DS128: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-DS128: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-DS128: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-DS128: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI-DS128: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-DS128: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-DS128: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI-DS128: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI-DS128: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-DS128: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-DS128: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI-DS128: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI-DS128: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; CI-DS128: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; CI-DS128: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI-DS128: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI-DS128: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; CI-DS128: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; CI-DS128: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; CI-DS128: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; CI-DS128: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; CI-DS128: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; CI-DS128: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; CI-DS128: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; CI-DS128: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; CI-DS128: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; VI-LABEL: name: test_load_local_v4s8_align4 ; VI-LABEL: name: test_load_local_v4s8_align4
; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<4 x s8>), addrspace 3) ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; VI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; VI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; GFX9-LABEL: name: test_load_local_v4s8_align4 ; GFX9-LABEL: name: test_load_local_v4s8_align4
; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<4 x s8>), addrspace 3) ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY1]](s32), [[COPY2]](s32) ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY1]](s32), [[COPY2]](s32)
; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[COPY4]](s32) ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[COPY4]](s32)
; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
; GFX9: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>) ; GFX9: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
; GFX9: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; GFX9: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; GFX9-UNALIGNED-LABEL: name: test_load_local_v4s8_align4 ; GFX9-UNALIGNED-LABEL: name: test_load_local_v4s8_align4
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<4 x s8>), addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; GFX9-UNALIGNED: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY1]](s32), [[COPY2]](s32) ; GFX9-UNALIGNED: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY1]](s32), [[COPY2]](s32)
; GFX9-UNALIGNED: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; GFX9-UNALIGNED: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; GFX9-UNALIGNED: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; GFX9-UNALIGNED: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; GFX9-UNALIGNED: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[COPY4]](s32) ; GFX9-UNALIGNED: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[COPY4]](s32)
; GFX9-UNALIGNED: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) ; GFX9-UNALIGNED: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
; GFX9-UNALIGNED: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>) ; GFX9-UNALIGNED: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
; GFX9-UNALIGNED: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; GFX9-UNALIGNED: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; GFX10-LABEL: name: test_load_local_v4s8_align4 ; GFX10-LABEL: name: test_load_local_v4s8_align4
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<4 x s8>), addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX10: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX10: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX10: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX10: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX10: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX10: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX10: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX10: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX10: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX10: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; GFX10: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; GFX10: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX10: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; GFX10: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY1]](s32), [[COPY2]](s32) ; GFX10: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY1]](s32), [[COPY2]](s32)
; GFX10: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; GFX10: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; GFX10: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; GFX10: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; GFX10: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[COPY4]](s32) ; GFX10: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[COPY4]](s32)
; GFX10: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) ; GFX10: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
; GFX10: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>) ; GFX10: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
; GFX10: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; GFX10: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; GFX10-UNALIGNED-LABEL: name: test_load_local_v4s8_align4 ; GFX10-UNALIGNED-LABEL: name: test_load_local_v4s8_align4
; GFX10-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<4 x s8>), addrspace 3) ; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
; GFX10-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX10-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX10-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX10-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX10-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX10-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX10-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX10-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX10-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX10-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX10-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX10-UNALIGNED: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; GFX10-UNALIGNED: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; GFX10-UNALIGNED: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX10-UNALIGNED: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
Show All 12 Lines
--- ---
name: test_load_local_v8s8_align8 name: test_load_local_v8s8_align8
body: | body: |
bb.0: bb.0:
liveins: $vgpr0 liveins: $vgpr0
; SI-LABEL: name: test_load_local_v8s8_align8 ; SI-LABEL: name: test_load_local_v8s8_align8
; SI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; SI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; SI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<8 x s8>), addrspace 3) ; SI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 10 Lines bb.0:
; SI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; SI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; SI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; SI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; SI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32) ; SI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32)
; SI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>) ; SI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>)
; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>) ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>)
; SI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; SI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>)
; CI-LABEL: name: test_load_local_v8s8_align8 ; CI-LABEL: name: test_load_local_v8s8_align8
; CI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; CI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; CI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<8 x s8>), addrspace 3) ; CI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
; CI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; CI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 10 Lines bb.0:
; CI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; CI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; CI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; CI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; CI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32) ; CI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32)
; CI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>) ; CI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>)
; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>) ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>)
; CI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; CI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>)
; CI-DS128-LABEL: name: test_load_local_v8s8_align8 ; CI-DS128-LABEL: name: test_load_local_v8s8_align8
; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; CI-DS128: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<8 x s8>), addrspace 3) ; CI-DS128: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
; CI-DS128: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; CI-DS128: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; CI-DS128: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-DS128: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-DS128: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; CI-DS128: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; CI-DS128: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI-DS128: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI-DS128: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; CI-DS128: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; CI-DS128: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI-DS128: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI-DS128: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; CI-DS128: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; CI-DS128: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; CI-DS128: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 10 Lines bb.0:
; CI-DS128: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; CI-DS128: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; CI-DS128: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; CI-DS128: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; CI-DS128: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32) ; CI-DS128: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32)
; CI-DS128: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>) ; CI-DS128: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>)
; CI-DS128: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>) ; CI-DS128: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>)
; CI-DS128: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; CI-DS128: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>)
; VI-LABEL: name: test_load_local_v8s8_align8 ; VI-LABEL: name: test_load_local_v8s8_align8
; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; VI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<8 x s8>), addrspace 3) ; VI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 10 Lines bb.0:
; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32)
; VI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; VI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; VI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32) ; VI: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32), [[COPY8]](s32)
; VI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>) ; VI: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR1]](<4 x s32>)
; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>) ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>)
; VI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; VI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>)
; GFX9-LABEL: name: test_load_local_v8s8_align8 ; GFX9-LABEL: name: test_load_local_v8s8_align8
; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<8 x s8>), addrspace 3) ; GFX9: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 14 Lines bb.0:
; GFX9: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; GFX9: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; GFX9: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY7]](s32), [[COPY8]](s32) ; GFX9: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY7]](s32), [[COPY8]](s32)
; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>) ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>)
; GFX9: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS1]](<4 x s16>) ; GFX9: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS1]](<4 x s16>)
; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>) ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>)
; GFX9: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS2]](<8 x s8>) ; GFX9: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS2]](<8 x s8>)
; GFX9-UNALIGNED-LABEL: name: test_load_local_v8s8_align8 ; GFX9-UNALIGNED-LABEL: name: test_load_local_v8s8_align8
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<8 x s8>), addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
; GFX9-UNALIGNED: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX9-UNALIGNED: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; GFX9-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; GFX9-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; GFX9-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; GFX9-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; GFX9-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 14 Lines bb.0:
; GFX9-UNALIGNED: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; GFX9-UNALIGNED: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; GFX9-UNALIGNED: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY7]](s32), [[COPY8]](s32) ; GFX9-UNALIGNED: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY7]](s32), [[COPY8]](s32)
; GFX9-UNALIGNED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>) ; GFX9-UNALIGNED: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>)
; GFX9-UNALIGNED: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS1]](<4 x s16>) ; GFX9-UNALIGNED: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS1]](<4 x s16>)
; GFX9-UNALIGNED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>) ; GFX9-UNALIGNED: [[CONCAT_VECTORS2:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>)
; GFX9-UNALIGNED: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS2]](<8 x s8>) ; GFX9-UNALIGNED: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS2]](<8 x s8>)
; GFX10-LABEL: name: test_load_local_v8s8_align8 ; GFX10-LABEL: name: test_load_local_v8s8_align8
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<8 x s8>), addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
; GFX10: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX10: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX10: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; GFX10: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; GFX10: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX10: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; GFX10: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; GFX10: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX10: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX10: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; GFX10: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; GFX10: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; GFX10: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
Show All 14 Lines bb.0:
; GFX10: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) ; GFX10: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32)
; GFX10: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY7]](s32), [[COPY8]](s32) ; GFX10: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY7]](s32), [[COPY8]](s32)
; GFX10: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>) ; GFX10: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>)
; GFX10: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS1]](<4 x s16>) ; GFX10: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS1]](<4 x s16>)
; GFX10: [[CONCAT_VECTORS2:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>) ; GFX10: [[CONCAT_VECTORS2:%[0-9]+]]:_(<8 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>)
; GFX10: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS2]](<8 x s8>) ; GFX10: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS2]](<8 x s8>)
; GFX10-UNALIGNED-LABEL: name: test_load_local_v8s8_align8 ; GFX10-UNALIGNED-LABEL: name: test_load_local_v8s8_align8
; GFX10-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<8 x s8>), addrspace 3) ; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
; GFX10-UNALIGNED: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX10-UNALIGNED: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
; GFX10-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX10-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX10-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; GFX10-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; GFX10-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX10-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; GFX10-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; GFX10-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX10-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX10-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; GFX10-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; GFX10-UNALIGNED: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; GFX10-UNALIGNED: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
▲ Show 20 Lines • Show All 356 Lines • ▼ Show 20 Lines bb.0:
; GFX9: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LOAD15]](s32) ; GFX9: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LOAD15]](s32)
; GFX9: [[BUILD_VECTOR_TRUNC7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY15]](s32), [[COPY16]](s32) ; GFX9: [[BUILD_VECTOR_TRUNC7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY15]](s32), [[COPY16]](s32)
; GFX9: [[CONCAT_VECTORS3:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC6]](<2 x s16>), [[BUILD_VECTOR_TRUNC7]](<2 x s16>) ; GFX9: [[CONCAT_VECTORS3:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC6]](<2 x s16>), [[BUILD_VECTOR_TRUNC7]](<2 x s16>)
; GFX9: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS3]](<4 x s16>) ; GFX9: [[TRUNC3:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS3]](<4 x s16>)
; GFX9: [[CONCAT_VECTORS4:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>) ; GFX9: [[CONCAT_VECTORS4:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC2]](<4 x s8>), [[TRUNC3]](<4 x s8>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS4]](<16 x s8>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS4]](<16 x s8>)
; GFX9-UNALIGNED-LABEL: name: test_load_local_v16s8_align16 ; GFX9-UNALIGNED-LABEL: name: test_load_local_v16s8_align16
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<16 x s8>), align 1, addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 1, addrspace 3)
; GFX9-UNALIGNED: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; GFX9-UNALIGNED: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; GFX9-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; GFX9-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) ; GFX9-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
; GFX9-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) ; GFX9-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
▲ Show 20 Lines • Show All 5,989 Lines • ▼ Show 20 Lines bb.0:
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
; GFX9: [[PTR_ADD2:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C2]](s32) ; GFX9: [[PTR_ADD2:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C2]](s32)
; GFX9: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s32) from unknown-address + 12, addrspace 3) ; GFX9: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s32) from unknown-address + 12, addrspace 3)
; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LOAD1]](s32), [[LOAD2]](s32), [[LOAD3]](s32) ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LOAD1]](s32), [[LOAD2]](s32), [[LOAD3]](s32)
; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX9-UNALIGNED-LABEL: name: test_load_local_v2p1_align4 ; GFX9-UNALIGNED-LABEL: name: test_load_local_v2p1_align4
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x p1>), align 4, addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 4, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
; GFX10-LABEL: name: test_load_local_v2p1_align4 ; GFX10-LABEL: name: test_load_local_v2p1_align4
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 4, addrspace 3) ; GFX10: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 4, addrspace 3)
▲ Show 20 Lines • Show All 1,454 Lines • ▼ Show 20 Lines bb.0:
; GFX9: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR11]](s32), [[OR14]](s32), [[OR17]](s32) ; GFX9: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR11]](s32), [[OR14]](s32), [[OR17]](s32)
; GFX9: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>) ; GFX9: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
; GFX9: [[COPY25:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9: [[COPY25:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; GFX9: [[COPY26:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; GFX9: [[COPY26:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[COPY25]](s96) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[COPY25]](s96)
; GFX9: $vgpr3_vgpr4_vgpr5 = COPY [[COPY26]](s96) ; GFX9: $vgpr3_vgpr4_vgpr5 = COPY [[COPY26]](s96)
; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s96_align1 ; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s96_align1
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 1, addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 1, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
; GFX9-UNALIGNED: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX9-UNALIGNED: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX9-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (s96) from unknown-address + 12, align 1, addrspace 3) ; GFX9-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<3 x s32>) from unknown-address + 12, align 1, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-UNALIGNED: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-UNALIGNED: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-UNALIGNED: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; GFX10-LABEL: name: test_load_local_v2s96_align1 ; GFX10-LABEL: name: test_load_local_v2s96_align1
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3)
▲ Show 20 Lines • Show All 680 Lines • ▼ Show 20 Lines bb.0:
; GFX9: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR3]](s32), [[OR4]](s32), [[OR5]](s32) ; GFX9: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR3]](s32), [[OR4]](s32), [[OR5]](s32)
; GFX9: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>) ; GFX9: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
; GFX9: [[COPY13:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9: [[COPY13:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; GFX9: [[COPY14:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; GFX9: [[COPY14:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[COPY13]](s96) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[COPY13]](s96)
; GFX9: $vgpr3_vgpr4_vgpr5 = COPY [[COPY14]](s96) ; GFX9: $vgpr3_vgpr4_vgpr5 = COPY [[COPY14]](s96)
; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s96_align2 ; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s96_align2
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 2, addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 2, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
; GFX9-UNALIGNED: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX9-UNALIGNED: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX9-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (s96) from unknown-address + 12, align 2, addrspace 3) ; GFX9-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<3 x s32>) from unknown-address + 12, align 2, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-UNALIGNED: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-UNALIGNED: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-UNALIGNED: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; GFX10-LABEL: name: test_load_local_v2s96_align2 ; GFX10-LABEL: name: test_load_local_v2s96_align2
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
▲ Show 20 Lines • Show All 260 Lines • ▼ Show 20 Lines bb.0:
; GFX9: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[UV3]](s32), [[LOAD3]](s32) ; GFX9: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[UV3]](s32), [[LOAD3]](s32)
; GFX9: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>) ; GFX9: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; GFX9: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; GFX9: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s96_align4 ; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s96_align4
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 4, addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 4, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
; GFX9-UNALIGNED: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX9-UNALIGNED: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX9-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (s96) from unknown-address + 12, align 4, addrspace 3) ; GFX9-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-UNALIGNED: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-UNALIGNED: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-UNALIGNED: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; GFX10-LABEL: name: test_load_local_v2s96_align4 ; GFX10-LABEL: name: test_load_local_v2s96_align4
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), align 4, addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), align 4, addrspace 3)
▲ Show 20 Lines • Show All 89 Lines • ▼ Show 20 Lines bb.0:
; CI: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[UV3]](s32), [[LOAD3]](s32) ; CI: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[UV3]](s32), [[LOAD3]](s32)
; CI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>) ; CI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; CI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; CI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; CI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; CI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; CI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; CI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; CI-DS128-LABEL: name: test_load_local_v2s96_align16 ; CI-DS128-LABEL: name: test_load_local_v2s96_align16
; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; CI-DS128: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 16, addrspace 3) ; CI-DS128: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 16, addrspace 3)
; CI-DS128: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-DS128: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; CI-DS128: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; CI-DS128: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
; CI-DS128: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; CI-DS128: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; CI-DS128: [[LOAD1:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<2 x s32>) from unknown-address + 12, align 4, addrspace 3) ; CI-DS128: [[LOAD1:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<2 x s32>) from unknown-address + 12, align 4, addrspace 3)
; CI-DS128: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-DS128: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI-DS128: [[PTR_ADD1:%[0-9]+]]:_(p3) = G_PTR_ADD [[PTR_ADD]], [[C1]](s32) ; CI-DS128: [[PTR_ADD1:%[0-9]+]]:_(p3) = G_PTR_ADD [[PTR_ADD]], [[C1]](s32)
; CI-DS128: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 20, addrspace 3) ; CI-DS128: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 20, addrspace 3)
; CI-DS128: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<2 x s32>) ; CI-DS128: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<2 x s32>)
; CI-DS128: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD2]](s32) ; CI-DS128: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD2]](s32)
; CI-DS128: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; CI-DS128: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; CI-DS128: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI-DS128: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; CI-DS128: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; CI-DS128: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; CI-DS128: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; CI-DS128: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; CI-DS128: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; CI-DS128: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; VI-LABEL: name: test_load_local_v2s96_align16 ; VI-LABEL: name: test_load_local_v2s96_align16
; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 16, addrspace 3) ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 16, addrspace 3)
; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
; VI: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; VI: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; VI: [[LOAD1:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<2 x s32>) from unknown-address + 12, align 4, addrspace 3) ; VI: [[LOAD1:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<2 x s32>) from unknown-address + 12, align 4, addrspace 3)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[PTR_ADD1:%[0-9]+]]:_(p3) = G_PTR_ADD [[PTR_ADD]], [[C1]](s32) ; VI: [[PTR_ADD1:%[0-9]+]]:_(p3) = G_PTR_ADD [[PTR_ADD]], [[C1]](s32)
; VI: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 20, addrspace 3) ; VI: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 20, addrspace 3)
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<2 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<2 x s32>)
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD2]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD2]](s32)
; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; VI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; VI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; VI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; GFX9-LABEL: name: test_load_local_v2s96_align16 ; GFX9-LABEL: name: test_load_local_v2s96_align16
; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 16, addrspace 3) ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 16, addrspace 3)
; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
; GFX9: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX9: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX9: [[LOAD1:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<2 x s32>) from unknown-address + 12, align 4, addrspace 3) ; GFX9: [[LOAD1:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<2 x s32>) from unknown-address + 12, align 4, addrspace 3)
; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9: [[PTR_ADD1:%[0-9]+]]:_(p3) = G_PTR_ADD [[PTR_ADD]], [[C1]](s32) ; GFX9: [[PTR_ADD1:%[0-9]+]]:_(p3) = G_PTR_ADD [[PTR_ADD]], [[C1]](s32)
; GFX9: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 20, addrspace 3) ; GFX9: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 20, addrspace 3)
; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<2 x s32>) ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<2 x s32>)
; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD2]](s32) ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD2]](s32)
; GFX9: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; GFX9: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; GFX9: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; GFX9: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s96_align16 ; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s96_align16
; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 16, addrspace 3) ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 16, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
; GFX9-UNALIGNED: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX9-UNALIGNED: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX9-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (s96) from unknown-address + 12, align 4, addrspace 3) ; GFX9-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 3)
; GFX9-UNALIGNED: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-UNALIGNED: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-UNALIGNED: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-UNALIGNED: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; GFX10-LABEL: name: test_load_local_v2s96_align16 ; GFX10-LABEL: name: test_load_local_v2s96_align16
; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 16, addrspace 3) ; GFX10: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 16, addrspace 3)
; GFX10: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX10: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX10: [[LOAD1:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<2 x s32>) from unknown-address + 12, align 4, addrspace 3) ; GFX10: [[LOAD1:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<2 x s32>) from unknown-address + 12, align 4, addrspace 3)
; GFX10: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX10: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX10: [[PTR_ADD1:%[0-9]+]]:_(p3) = G_PTR_ADD [[PTR_ADD]], [[C1]](s32) ; GFX10: [[PTR_ADD1:%[0-9]+]]:_(p3) = G_PTR_ADD [[PTR_ADD]], [[C1]](s32)
; GFX10: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 20, addrspace 3) ; GFX10: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 20, addrspace 3)
; GFX10: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<2 x s32>) ; GFX10: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<2 x s32>)
; GFX10: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD2]](s32) ; GFX10: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD2]](s32)
; GFX10: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>) ; GFX10: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX10: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX10: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
; GFX10: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) ; GFX10: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX10: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96) ; GFX10: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX10: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX10: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
; GFX10-UNALIGNED-LABEL: name: test_load_local_v2s96_align16 ; GFX10-UNALIGNED-LABEL: name: test_load_local_v2s96_align16
; GFX10-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX10-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 16, addrspace 3) ; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 16, addrspace 3)
; GFX10-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX10-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX10-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX10-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
; GFX10-UNALIGNED: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) ; GFX10-UNALIGNED: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
; GFX10-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<2 x s32>) from unknown-address + 12, align 4, addrspace 3) ; GFX10-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<2 x s32>) from unknown-address + 12, align 4, addrspace 3)
; GFX10-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX10-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX10-UNALIGNED: [[PTR_ADD1:%[0-9]+]]:_(p3) = G_PTR_ADD [[PTR_ADD]], [[C1]](s32) ; GFX10-UNALIGNED: [[PTR_ADD1:%[0-9]+]]:_(p3) = G_PTR_ADD [[PTR_ADD]], [[C1]](s32)
; GFX10-UNALIGNED: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 20, addrspace 3) ; GFX10-UNALIGNED: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 20, addrspace 3)
; GFX10-UNALIGNED: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<2 x s32>) ; GFX10-UNALIGNED: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<2 x s32>)
Show All 13 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-private.mir

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 4,520 Lines • ▼ Show 20 Lines
--- ---
name: test_load_private_v2s8_align2 name: test_load_private_v2s8_align2
body: | body: |
bb.0: bb.0:
liveins: $vgpr0 liveins: $vgpr0
; SI-LABEL: name: test_load_private_v2s8_align2 ; SI-LABEL: name: test_load_private_v2s8_align2
; SI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 ; SI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (<2 x s8>), addrspace 5) ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; SI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; SI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; SI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; SI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32) ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
; SI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; SI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]] ; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]]
; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32) ; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; SI: $vgpr0 = COPY [[ANYEXT]](s32) ; SI: $vgpr0 = COPY [[ANYEXT]](s32)
; CI-LABEL: name: test_load_private_v2s8_align2 ; CI-LABEL: name: test_load_private_v2s8_align2
; CI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 ; CI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (<2 x s8>), addrspace 5) ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; CI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; CI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; CI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; CI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; CI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32) ; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C]](s32)
; CI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; CI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]] ; CI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C4]]
; CI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32) ; CI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY1]](s32)
; CI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; CI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; CI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; CI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; CI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; CI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; CI: $vgpr0 = COPY [[ANYEXT]](s32) ; CI: $vgpr0 = COPY [[ANYEXT]](s32)
; VI-LABEL: name: test_load_private_v2s8_align2 ; VI-LABEL: name: test_load_private_v2s8_align2
; VI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 ; VI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (<2 x s8>), addrspace 5) ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; VI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; VI: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] ; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]] ; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
; VI: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 ; VI: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16) ; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; VI: $vgpr0 = COPY [[ANYEXT]](s32)
; GFX9-LABEL: name: test_load_private_v2s8_align2 ; GFX9-LABEL: name: test_load_private_v2s8_align2
; GFX9: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 ; GFX9: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (<2 x s8>), addrspace 5) ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; GFX9: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
▲ Show 20 Lines • Show All 253 Lines • ▼ Show 20 Lines
--- ---
name: test_load_private_v4s8_align4 name: test_load_private_v4s8_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0 liveins: $vgpr0
; SI-LABEL: name: test_load_private_v4s8_align4 ; SI-LABEL: name: test_load_private_v4s8_align4
; SI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 ; SI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (<4 x s8>), addrspace 5) ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5)
; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; SI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; SI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; SI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; SI: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; CI-LABEL: name: test_load_private_v4s8_align4 ; CI-LABEL: name: test_load_private_v4s8_align4
; CI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 ; CI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (<4 x s8>), addrspace 5) ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5)
; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; CI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; CI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; CI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; CI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; CI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; CI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; CI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; CI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; CI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; CI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; CI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; CI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; CI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; CI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; CI: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; VI-LABEL: name: test_load_private_v4s8_align4 ; VI-LABEL: name: test_load_private_v4s8_align4
; VI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 ; VI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (<4 x s8>), addrspace 5) ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; VI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32) ; VI: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32)
; VI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>) ; VI: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[BUILD_VECTOR]](<4 x s32>)
; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>)
; GFX9-LABEL: name: test_load_private_v4s8_align4 ; GFX9-LABEL: name: test_load_private_v4s8_align4
; GFX9: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 ; GFX9: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (<4 x s8>), addrspace 5) ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5)
; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LOAD]](s32)
; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
▲ Show 20 Lines • Show All 6,375 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store-global.mir

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 3,612 Lines • ▼ Show 20 Lines bb.0:
; SI: G_STORE [[COPY16]](s32), [[PTR_ADD13]](p1) :: (store (s8) into unknown-address + 14, addrspace 1) ; SI: G_STORE [[COPY16]](s32), [[PTR_ADD13]](p1) :: (store (s8) into unknown-address + 14, addrspace 1)
; SI: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD11]], [[C5]](s64) ; SI: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD11]], [[C5]](s64)
; SI: [[COPY17:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32) ; SI: [[COPY17:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32)
; SI: G_STORE [[COPY17]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1) ; SI: G_STORE [[COPY17]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1)
; CI-LABEL: name: test_store_global_v8s16_align1 ; CI-LABEL: name: test_store_global_v8s16_align1
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 1, addrspace 1) ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
; VI-LABEL: name: test_store_global_v8s16_align1 ; VI-LABEL: name: test_store_global_v8s16_align1
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<4 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<4 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
▲ Show 20 Lines • Show All 61 Lines • ▼ Show 20 Lines bb.0:
; VI: G_STORE [[COPY16]](s32), [[PTR_ADD13]](p1) :: (store (s8) into unknown-address + 14, addrspace 1) ; VI: G_STORE [[COPY16]](s32), [[PTR_ADD13]](p1) :: (store (s8) into unknown-address + 14, addrspace 1)
; VI: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD11]], [[C5]](s64) ; VI: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD11]], [[C5]](s64)
; VI: [[COPY17:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32) ; VI: [[COPY17:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32)
; VI: G_STORE [[COPY17]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1) ; VI: G_STORE [[COPY17]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1)
; GFX9-LABEL: name: test_store_global_v8s16_align1 ; GFX9-LABEL: name: test_store_global_v8s16_align1
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 1, addrspace 1) ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<8 x s16>), align 1, addrspace 1) G_STORE %1, %0 :: (store (<8 x s16>), align 1, addrspace 1)
... ...
--- ---
name: test_store_global_v8s16_align2 name: test_store_global_v8s16_align2
body: | body: |
Show All 37 Lines bb.0:
; SI: G_STORE [[COPY8]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1) ; SI: G_STORE [[COPY8]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1)
; SI: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64) ; SI: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64)
; SI: [[COPY9:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32) ; SI: [[COPY9:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32)
; SI: G_STORE [[COPY9]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1) ; SI: G_STORE [[COPY9]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1)
; CI-LABEL: name: test_store_global_v8s16_align2 ; CI-LABEL: name: test_store_global_v8s16_align2
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 2, addrspace 1) ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
; VI-LABEL: name: test_store_global_v8s16_align2 ; VI-LABEL: name: test_store_global_v8s16_align2
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; VI: [[UV:%[0-9]+]]:_(<2 x s32>), [[UV1:%[0-9]+]]:_(<2 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<4 x s32>) ; VI: [[UV:%[0-9]+]]:_(<2 x s32>), [[UV1:%[0-9]+]]:_(<2 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<4 x s32>)
; VI: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV]](<2 x s32>) ; VI: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV]](<2 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C]](s32)
Show All 26 Lines bb.0:
; VI: G_STORE [[COPY8]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1) ; VI: G_STORE [[COPY8]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1)
; VI: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64) ; VI: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64)
; VI: [[COPY9:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32) ; VI: [[COPY9:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32)
; VI: G_STORE [[COPY9]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1) ; VI: G_STORE [[COPY9]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1)
; GFX9-LABEL: name: test_store_global_v8s16_align2 ; GFX9-LABEL: name: test_store_global_v8s16_align2
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 2, addrspace 1) ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<8 x s16>), align 2, addrspace 1) G_STORE %1, %0 :: (store (<8 x s16>), align 2, addrspace 1)
... ...
--- ---
name: test_store_global_v8s16_align4 name: test_store_global_v8s16_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5 liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; SI-LABEL: name: test_store_global_v8s16_align4 ; SI-LABEL: name: test_store_global_v8s16_align4
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 4, addrspace 1) ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
; CI-LABEL: name: test_store_global_v8s16_align4 ; CI-LABEL: name: test_store_global_v8s16_align4
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 4, addrspace 1) ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
; VI-LABEL: name: test_store_global_v8s16_align4 ; VI-LABEL: name: test_store_global_v8s16_align4
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 4, addrspace 1) ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
; GFX9-LABEL: name: test_store_global_v8s16_align4 ; GFX9-LABEL: name: test_store_global_v8s16_align4
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 4, addrspace 1) ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<8 x s16>), align 4, addrspace 1) G_STORE %1, %0 :: (store (<8 x s16>), align 4, addrspace 1)
... ...
--- ---
name: test_store_global_v8s16_align8 name: test_store_global_v8s16_align8
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5 liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; SI-LABEL: name: test_store_global_v8s16_align8 ; SI-LABEL: name: test_store_global_v8s16_align8
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 8, addrspace 1) ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
; CI-LABEL: name: test_store_global_v8s16_align8 ; CI-LABEL: name: test_store_global_v8s16_align8
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 8, addrspace 1) ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
; VI-LABEL: name: test_store_global_v8s16_align8 ; VI-LABEL: name: test_store_global_v8s16_align8
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 8, addrspace 1) ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
; GFX9-LABEL: name: test_store_global_v8s16_align8 ; GFX9-LABEL: name: test_store_global_v8s16_align8
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 8, addrspace 1) ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<8 x s16>), align 8, addrspace 1) G_STORE %1, %0 :: (store (<8 x s16>), align 8, addrspace 1)
... ...
--- ---
name: test_store_global_v8s16_align16 name: test_store_global_v8s16_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5 liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; SI-LABEL: name: test_store_global_v8s16_align16 ; SI-LABEL: name: test_store_global_v8s16_align16
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), addrspace 1) ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
; CI-LABEL: name: test_store_global_v8s16_align16 ; CI-LABEL: name: test_store_global_v8s16_align16
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), addrspace 1) ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
; VI-LABEL: name: test_store_global_v8s16_align16 ; VI-LABEL: name: test_store_global_v8s16_align16
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), addrspace 1) ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
; GFX9-LABEL: name: test_store_global_v8s16_align16 ; GFX9-LABEL: name: test_store_global_v8s16_align16
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), addrspace 1) ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<8 x s16>), align 16, addrspace 1) G_STORE %1, %0 :: (store (<8 x s16>), align 16, addrspace 1)
... ...
--- ---
name: test_store_global_v2p0_align1 name: test_store_global_v2p0_align1
body: | body: |
▲ Show 20 Lines • Show All 72 Lines • ▼ Show 20 Lines bb.0:
; SI: G_STORE [[COPY16]](s32), [[PTR_ADD13]](p1) :: (store (s8) into unknown-address + 14, addrspace 1) ; SI: G_STORE [[COPY16]](s32), [[PTR_ADD13]](p1) :: (store (s8) into unknown-address + 14, addrspace 1)
; SI: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD11]], [[C5]](s64) ; SI: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD11]], [[C5]](s64)
; SI: [[COPY17:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32) ; SI: [[COPY17:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32)
; SI: G_STORE [[COPY17]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1) ; SI: G_STORE [[COPY17]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1)
; CI-LABEL: name: test_store_global_v2p0_align1 ; CI-LABEL: name: test_store_global_v2p0_align1
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 1, addrspace 1) ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
; VI-LABEL: name: test_store_global_v2p0_align1 ; VI-LABEL: name: test_store_global_v2p0_align1
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<4 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<4 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
▲ Show 20 Lines • Show All 61 Lines • ▼ Show 20 Lines bb.0:
; VI: G_STORE [[COPY16]](s32), [[PTR_ADD13]](p1) :: (store (s8) into unknown-address + 14, addrspace 1) ; VI: G_STORE [[COPY16]](s32), [[PTR_ADD13]](p1) :: (store (s8) into unknown-address + 14, addrspace 1)
; VI: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD11]], [[C5]](s64) ; VI: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD11]], [[C5]](s64)
; VI: [[COPY17:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32) ; VI: [[COPY17:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32)
; VI: G_STORE [[COPY17]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1) ; VI: G_STORE [[COPY17]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1)
; GFX9-LABEL: name: test_store_global_v2p0_align1 ; GFX9-LABEL: name: test_store_global_v2p0_align1
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 1, addrspace 1) ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<2 x p0>), align 1, addrspace 1) G_STORE %1, %0 :: (store (<2 x p0>), align 1, addrspace 1)
... ...
--- ---
name: test_store_global_v2p0_align2 name: test_store_global_v2p0_align2
body: | body: |
Show All 37 Lines bb.0:
; SI: G_STORE [[COPY8]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1) ; SI: G_STORE [[COPY8]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1)
; SI: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64) ; SI: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64)
; SI: [[COPY9:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32) ; SI: [[COPY9:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32)
; SI: G_STORE [[COPY9]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1) ; SI: G_STORE [[COPY9]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1)
; CI-LABEL: name: test_store_global_v2p0_align2 ; CI-LABEL: name: test_store_global_v2p0_align2
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 2, addrspace 1) ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
; VI-LABEL: name: test_store_global_v2p0_align2 ; VI-LABEL: name: test_store_global_v2p0_align2
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; VI: [[UV:%[0-9]+]]:_(<2 x s32>), [[UV1:%[0-9]+]]:_(<2 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<4 x s32>) ; VI: [[UV:%[0-9]+]]:_(<2 x s32>), [[UV1:%[0-9]+]]:_(<2 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<4 x s32>)
; VI: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV]](<2 x s32>) ; VI: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV]](<2 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C]](s32)
Show All 26 Lines bb.0:
; VI: G_STORE [[COPY8]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1) ; VI: G_STORE [[COPY8]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1)
; VI: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64) ; VI: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64)
; VI: [[COPY9:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32) ; VI: [[COPY9:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32)
; VI: G_STORE [[COPY9]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1) ; VI: G_STORE [[COPY9]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1)
; GFX9-LABEL: name: test_store_global_v2p0_align2 ; GFX9-LABEL: name: test_store_global_v2p0_align2
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 2, addrspace 1) ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<2 x p0>), align 2, addrspace 1) G_STORE %1, %0 :: (store (<2 x p0>), align 2, addrspace 1)
... ...
--- ---
name: test_store_global_v2p0_align4 name: test_store_global_v2p0_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5 liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; SI-LABEL: name: test_store_global_v2p0_align4 ; SI-LABEL: name: test_store_global_v2p0_align4
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 4, addrspace 1) ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
; CI-LABEL: name: test_store_global_v2p0_align4 ; CI-LABEL: name: test_store_global_v2p0_align4
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 4, addrspace 1) ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
; VI-LABEL: name: test_store_global_v2p0_align4 ; VI-LABEL: name: test_store_global_v2p0_align4
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 4, addrspace 1) ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
; GFX9-LABEL: name: test_store_global_v2p0_align4 ; GFX9-LABEL: name: test_store_global_v2p0_align4
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 4, addrspace 1) ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<2 x p0>), align 4, addrspace 1) G_STORE %1, %0 :: (store (<2 x p0>), align 4, addrspace 1)
... ...
--- ---
name: test_store_global_v2p0_align8 name: test_store_global_v2p0_align8
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5 liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; SI-LABEL: name: test_store_global_v2p0_align8 ; SI-LABEL: name: test_store_global_v2p0_align8
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 8, addrspace 1) ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
; CI-LABEL: name: test_store_global_v2p0_align8 ; CI-LABEL: name: test_store_global_v2p0_align8
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 8, addrspace 1) ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
; VI-LABEL: name: test_store_global_v2p0_align8 ; VI-LABEL: name: test_store_global_v2p0_align8
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 8, addrspace 1) ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
; GFX9-LABEL: name: test_store_global_v2p0_align8 ; GFX9-LABEL: name: test_store_global_v2p0_align8
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 8, addrspace 1) ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<2 x p0>), align 8, addrspace 1) G_STORE %1, %0 :: (store (<2 x p0>), align 8, addrspace 1)
... ...
--- ---
name: test_store_global_v2p0_align16 name: test_store_global_v2p0_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5 liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; SI-LABEL: name: test_store_global_v2p0_align16 ; SI-LABEL: name: test_store_global_v2p0_align16
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), addrspace 1) ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
; CI-LABEL: name: test_store_global_v2p0_align16 ; CI-LABEL: name: test_store_global_v2p0_align16
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), addrspace 1) ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
; VI-LABEL: name: test_store_global_v2p0_align16 ; VI-LABEL: name: test_store_global_v2p0_align16
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), addrspace 1) ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
; GFX9-LABEL: name: test_store_global_v2p0_align16 ; GFX9-LABEL: name: test_store_global_v2p0_align16
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), addrspace 1) ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<2 x p0>), align 16, addrspace 1) G_STORE %1, %0 :: (store (<2 x p0>), align 16, addrspace 1)
... ...
--- ---
name: test_store_global_s96_align1 name: test_store_global_s96_align1
body: | body: |
▲ Show 20 Lines • Show All 58 Lines • ▼ Show 20 Lines bb.0:
; SI: G_STORE [[COPY12]](s32), [[PTR_ADD9]](p1) :: (store (s8) into unknown-address + 10, addrspace 1) ; SI: G_STORE [[COPY12]](s32), [[PTR_ADD9]](p1) :: (store (s8) into unknown-address + 10, addrspace 1)
; SI: [[PTR_ADD10:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C5]](s64) ; SI: [[PTR_ADD10:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C5]](s64)
; SI: [[COPY13:%[0-9]+]]:_(s32) = COPY [[LSHR8]](s32) ; SI: [[COPY13:%[0-9]+]]:_(s32) = COPY [[LSHR8]](s32)
; SI: G_STORE [[COPY13]](s32), [[PTR_ADD10]](p1) :: (store (s8) into unknown-address + 11, addrspace 1) ; SI: G_STORE [[COPY13]](s32), [[PTR_ADD10]](p1) :: (store (s8) into unknown-address + 11, addrspace 1)
; CI-LABEL: name: test_store_global_s96_align1 ; CI-LABEL: name: test_store_global_s96_align1
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; CI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) ; CI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 1, addrspace 1) ; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 1, addrspace 1)
; VI-LABEL: name: test_store_global_s96_align1 ; VI-LABEL: name: test_store_global_s96_align1
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; VI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) ; VI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; VI: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[BITCAST]](<3 x s32>), 0 ; VI: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[BITCAST]](<3 x s32>), 0
; VI: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<3 x s32>), 64 ; VI: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<3 x s32>), 64
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<2 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<2 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
▲ Show 20 Lines • Show All 47 Lines • ▼ Show 20 Lines bb.0:
; VI: G_STORE [[COPY12]](s32), [[PTR_ADD9]](p1) :: (store (s8) into unknown-address + 10, addrspace 1) ; VI: G_STORE [[COPY12]](s32), [[PTR_ADD9]](p1) :: (store (s8) into unknown-address + 10, addrspace 1)
; VI: [[PTR_ADD10:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C5]](s64) ; VI: [[PTR_ADD10:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C5]](s64)
; VI: [[COPY13:%[0-9]+]]:_(s32) = COPY [[LSHR8]](s32) ; VI: [[COPY13:%[0-9]+]]:_(s32) = COPY [[LSHR8]](s32)
; VI: G_STORE [[COPY13]](s32), [[PTR_ADD10]](p1) :: (store (s8) into unknown-address + 11, addrspace 1) ; VI: G_STORE [[COPY13]](s32), [[PTR_ADD10]](p1) :: (store (s8) into unknown-address + 11, addrspace 1)
; GFX9-LABEL: name: test_store_global_s96_align1 ; GFX9-LABEL: name: test_store_global_s96_align1
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; GFX9: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) ; GFX9: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 1, addrspace 1) ; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = COPY $vgpr2_vgpr3_vgpr4 %1:_(s96) = COPY $vgpr2_vgpr3_vgpr4
G_STORE %1, %0 :: (store (s96), align 1, addrspace 1) G_STORE %1, %0 :: (store (s96), align 1, addrspace 1)
... ...
--- ---
name: test_store_global_s96_align2 name: test_store_global_s96_align2
body: | body: |
Show All 30 Lines bb.0:
; SI: G_STORE [[COPY6]](s32), [[PTR_ADD3]](p1) :: (store (s16) into unknown-address + 8, addrspace 1) ; SI: G_STORE [[COPY6]](s32), [[PTR_ADD3]](p1) :: (store (s16) into unknown-address + 8, addrspace 1)
; SI: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD3]], [[C1]](s64) ; SI: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD3]], [[C1]](s64)
; SI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; SI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; SI: G_STORE [[COPY7]](s32), [[PTR_ADD4]](p1) :: (store (s16) into unknown-address + 10, addrspace 1) ; SI: G_STORE [[COPY7]](s32), [[PTR_ADD4]](p1) :: (store (s16) into unknown-address + 10, addrspace 1)
; CI-LABEL: name: test_store_global_s96_align2 ; CI-LABEL: name: test_store_global_s96_align2
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; CI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) ; CI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 2, addrspace 1) ; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 2, addrspace 1)
; VI-LABEL: name: test_store_global_s96_align2 ; VI-LABEL: name: test_store_global_s96_align2
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; VI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) ; VI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; VI: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[BITCAST]](<3 x s32>), 0 ; VI: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[BITCAST]](<3 x s32>), 0
; VI: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<3 x s32>), 64 ; VI: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<3 x s32>), 64
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<2 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<2 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
Show All 19 Lines bb.0:
; VI: G_STORE [[COPY6]](s32), [[PTR_ADD3]](p1) :: (store (s16) into unknown-address + 8, addrspace 1) ; VI: G_STORE [[COPY6]](s32), [[PTR_ADD3]](p1) :: (store (s16) into unknown-address + 8, addrspace 1)
; VI: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD3]], [[C1]](s64) ; VI: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD3]], [[C1]](s64)
; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32)
; VI: G_STORE [[COPY7]](s32), [[PTR_ADD4]](p1) :: (store (s16) into unknown-address + 10, addrspace 1) ; VI: G_STORE [[COPY7]](s32), [[PTR_ADD4]](p1) :: (store (s16) into unknown-address + 10, addrspace 1)
; GFX9-LABEL: name: test_store_global_s96_align2 ; GFX9-LABEL: name: test_store_global_s96_align2
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; GFX9: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) ; GFX9: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 2, addrspace 1) ; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = COPY $vgpr2_vgpr3_vgpr4 %1:_(s96) = COPY $vgpr2_vgpr3_vgpr4
G_STORE %1, %0 :: (store (s96), align 2, addrspace 1) G_STORE %1, %0 :: (store (s96), align 2, addrspace 1)
... ...
--- ---
name: test_store_global_s96_align4 name: test_store_global_s96_align4
body: | body: |
Show All 9 Lines bb.0:
; SI: G_STORE [[EXTRACT]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1) ; SI: G_STORE [[EXTRACT]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
; SI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8 ; SI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
; SI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; SI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; SI: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, addrspace 1) ; SI: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, addrspace 1)
; CI-LABEL: name: test_store_global_s96_align4 ; CI-LABEL: name: test_store_global_s96_align4
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; CI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) ; CI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 4, addrspace 1) ; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
; VI-LABEL: name: test_store_global_s96_align4 ; VI-LABEL: name: test_store_global_s96_align4
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; VI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) ; VI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; VI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 4, addrspace 1) ; VI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
; GFX9-LABEL: name: test_store_global_s96_align4 ; GFX9-LABEL: name: test_store_global_s96_align4
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; GFX9: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) ; GFX9: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 4, addrspace 1) ; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = COPY $vgpr2_vgpr3_vgpr4 %1:_(s96) = COPY $vgpr2_vgpr3_vgpr4
G_STORE %1, %0 :: (store (s96), align 4, addrspace 1) G_STORE %1, %0 :: (store (s96), align 4, addrspace 1)
... ...
--- ---
name: test_store_global_s96_align8 name: test_store_global_s96_align8
body: | body: |
Show All 9 Lines bb.0:
; SI: G_STORE [[EXTRACT]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1) ; SI: G_STORE [[EXTRACT]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1)
; SI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8 ; SI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
; SI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; SI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; SI: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, align 8, addrspace 1) ; SI: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, align 8, addrspace 1)
; CI-LABEL: name: test_store_global_s96_align8 ; CI-LABEL: name: test_store_global_s96_align8
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; CI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) ; CI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 8, addrspace 1) ; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)
; VI-LABEL: name: test_store_global_s96_align8 ; VI-LABEL: name: test_store_global_s96_align8
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; VI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) ; VI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; VI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 8, addrspace 1) ; VI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)
; GFX9-LABEL: name: test_store_global_s96_align8 ; GFX9-LABEL: name: test_store_global_s96_align8
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; GFX9: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) ; GFX9: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 8, addrspace 1) ; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = COPY $vgpr2_vgpr3_vgpr4 %1:_(s96) = COPY $vgpr2_vgpr3_vgpr4
G_STORE %1, %0 :: (store (s96), align 8, addrspace 1) G_STORE %1, %0 :: (store (s96), align 8, addrspace 1)
... ...
--- ---
name: test_store_global_s96_align16 name: test_store_global_s96_align16
body: | body: |
Show All 9 Lines bb.0:
; SI: G_STORE [[EXTRACT]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1) ; SI: G_STORE [[EXTRACT]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
; SI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8 ; SI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
; SI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; SI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
; SI: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, align 8, addrspace 1) ; SI: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, align 8, addrspace 1)
; CI-LABEL: name: test_store_global_s96_align16 ; CI-LABEL: name: test_store_global_s96_align16
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; CI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) ; CI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 16, addrspace 1) ; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
; VI-LABEL: name: test_store_global_s96_align16 ; VI-LABEL: name: test_store_global_s96_align16
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; VI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) ; VI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; VI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 16, addrspace 1) ; VI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
; GFX9-LABEL: name: test_store_global_s96_align16 ; GFX9-LABEL: name: test_store_global_s96_align16
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; GFX9: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) ; GFX9: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 16, addrspace 1) ; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = COPY $vgpr2_vgpr3_vgpr4 %1:_(s96) = COPY $vgpr2_vgpr3_vgpr4
G_STORE %1, %0 :: (store (s96), align 16, addrspace 1) G_STORE %1, %0 :: (store (s96), align 16, addrspace 1)
... ...
--- ---
name: test_store_global_s128_align1 name: test_store_global_s128_align1
body: | body: |
▲ Show 20 Lines • Show All 72 Lines • ▼ Show 20 Lines bb.0:
; SI: G_STORE [[COPY16]](s32), [[PTR_ADD13]](p1) :: (store (s8) into unknown-address + 14, addrspace 1) ; SI: G_STORE [[COPY16]](s32), [[PTR_ADD13]](p1) :: (store (s8) into unknown-address + 14, addrspace 1)
; SI: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD11]], [[C5]](s64) ; SI: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD11]], [[C5]](s64)
; SI: [[COPY17:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32) ; SI: [[COPY17:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32)
; SI: G_STORE [[COPY17]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1) ; SI: G_STORE [[COPY17]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1)
; CI-LABEL: name: test_store_global_s128_align1 ; CI-LABEL: name: test_store_global_s128_align1
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 1, addrspace 1) ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
; VI-LABEL: name: test_store_global_s128_align1 ; VI-LABEL: name: test_store_global_s128_align1
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<4 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<4 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
▲ Show 20 Lines • Show All 61 Lines • ▼ Show 20 Lines bb.0:
; VI: G_STORE [[COPY16]](s32), [[PTR_ADD13]](p1) :: (store (s8) into unknown-address + 14, addrspace 1) ; VI: G_STORE [[COPY16]](s32), [[PTR_ADD13]](p1) :: (store (s8) into unknown-address + 14, addrspace 1)
; VI: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD11]], [[C5]](s64) ; VI: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD11]], [[C5]](s64)
; VI: [[COPY17:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32) ; VI: [[COPY17:%[0-9]+]]:_(s32) = COPY [[LSHR11]](s32)
; VI: G_STORE [[COPY17]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1) ; VI: G_STORE [[COPY17]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1)
; GFX9-LABEL: name: test_store_global_s128_align1 ; GFX9-LABEL: name: test_store_global_s128_align1
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 1, addrspace 1) ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (s128), align 1, addrspace 1) G_STORE %1, %0 :: (store (s128), align 1, addrspace 1)
... ...
--- ---
name: test_store_global_s128_align2 name: test_store_global_s128_align2
body: | body: |
Show All 37 Lines bb.0:
; SI: G_STORE [[COPY8]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1) ; SI: G_STORE [[COPY8]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1)
; SI: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64) ; SI: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64)
; SI: [[COPY9:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32) ; SI: [[COPY9:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32)
; SI: G_STORE [[COPY9]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1) ; SI: G_STORE [[COPY9]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1)
; CI-LABEL: name: test_store_global_s128_align2 ; CI-LABEL: name: test_store_global_s128_align2
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 2, addrspace 1) ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
; VI-LABEL: name: test_store_global_s128_align2 ; VI-LABEL: name: test_store_global_s128_align2
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; VI: [[UV:%[0-9]+]]:_(<2 x s32>), [[UV1:%[0-9]+]]:_(<2 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<4 x s32>) ; VI: [[UV:%[0-9]+]]:_(<2 x s32>), [[UV1:%[0-9]+]]:_(<2 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<4 x s32>)
; VI: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV]](<2 x s32>) ; VI: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV]](<2 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C]](s32) ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C]](s32)
Show All 26 Lines bb.0:
; VI: G_STORE [[COPY8]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1) ; VI: G_STORE [[COPY8]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1)
; VI: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64) ; VI: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64)
; VI: [[COPY9:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32) ; VI: [[COPY9:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32)
; VI: G_STORE [[COPY9]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1) ; VI: G_STORE [[COPY9]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1)
; GFX9-LABEL: name: test_store_global_s128_align2 ; GFX9-LABEL: name: test_store_global_s128_align2
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 2, addrspace 1) ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (s128), align 2, addrspace 1) G_STORE %1, %0 :: (store (s128), align 2, addrspace 1)
... ...
--- ---
name: test_store_global_s128_align4 name: test_store_global_s128_align4
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5 liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; SI-LABEL: name: test_store_global_s128_align4 ; SI-LABEL: name: test_store_global_s128_align4
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 4, addrspace 1) ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
; CI-LABEL: name: test_store_global_s128_align4 ; CI-LABEL: name: test_store_global_s128_align4
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 4, addrspace 1) ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
; VI-LABEL: name: test_store_global_s128_align4 ; VI-LABEL: name: test_store_global_s128_align4
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 4, addrspace 1) ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
; GFX9-LABEL: name: test_store_global_s128_align4 ; GFX9-LABEL: name: test_store_global_s128_align4
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 4, addrspace 1) ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (s128), align 4, addrspace 1) G_STORE %1, %0 :: (store (s128), align 4, addrspace 1)
... ...
--- ---
name: test_store_global_s128_align8 name: test_store_global_s128_align8
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5 liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; SI-LABEL: name: test_store_global_s128_align8 ; SI-LABEL: name: test_store_global_s128_align8
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 8, addrspace 1) ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
; CI-LABEL: name: test_store_global_s128_align8 ; CI-LABEL: name: test_store_global_s128_align8
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 8, addrspace 1) ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
; VI-LABEL: name: test_store_global_s128_align8 ; VI-LABEL: name: test_store_global_s128_align8
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 8, addrspace 1) ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
; GFX9-LABEL: name: test_store_global_s128_align8 ; GFX9-LABEL: name: test_store_global_s128_align8
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 8, addrspace 1) ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (s128), align 8, addrspace 1) G_STORE %1, %0 :: (store (s128), align 8, addrspace 1)
... ...
--- ---
name: test_store_global_s128_align16 name: test_store_global_s128_align16
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5 liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; SI-LABEL: name: test_store_global_s128_align16 ; SI-LABEL: name: test_store_global_s128_align16
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), addrspace 1) ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
; CI-LABEL: name: test_store_global_s128_align16 ; CI-LABEL: name: test_store_global_s128_align16
; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; CI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), addrspace 1) ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
; VI-LABEL: name: test_store_global_s128_align16 ; VI-LABEL: name: test_store_global_s128_align16
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), addrspace 1) ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
; GFX9-LABEL: name: test_store_global_s128_align16 ; GFX9-LABEL: name: test_store_global_s128_align16
; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), addrspace 1) ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (s128), align 16, addrspace 1) G_STORE %1, %0 :: (store (s128), align 16, addrspace 1)
... ...
--- ---
name: test_store_global_v5s32_align1 name: test_store_global_v5s32_align1
body: | body: |
▲ Show 20 Lines • Show All 3,750 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store.mir

Show First 20 Lines • Show All 246 Lines • ▼ Show 20 Lines
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5 liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; SI-LABEL: name: test_truncstore_global_s128_to_s8 ; SI-LABEL: name: test_truncstore_global_s128_to_s8
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), addrspace 1) ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
; VI-LABEL: name: test_truncstore_global_s128_to_s8 ; VI-LABEL: name: test_truncstore_global_s128_to_s8
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), addrspace 1) ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (s128), addrspace 1) G_STORE %1, %0 :: (store (s128), addrspace 1)
... ...
--- ---
name: test_store_global_i1 name: test_store_global_i1
body: | body: |
▲ Show 20 Lines • Show All 81 Lines • ▼ Show 20 Lines bb.0:
; SI: G_STORE [[EXTRACT]](<2 x s32>), [[COPY1]](p1) :: (store (<2 x s32>), align 16, addrspace 1) ; SI: G_STORE [[EXTRACT]](<2 x s32>), [[COPY1]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
; SI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8 ; SI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
; SI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY1]], [[C]](s64) ; SI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY1]], [[C]](s64)
; SI: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, align 8, addrspace 1) ; SI: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, align 8, addrspace 1)
; VI-LABEL: name: test_store_global_96 ; VI-LABEL: name: test_store_global_96
; VI: [[COPY:%[0-9]+]]:_(s96) = COPY $vgpr0_vgpr1_vgpr2 ; VI: [[COPY:%[0-9]+]]:_(s96) = COPY $vgpr0_vgpr1_vgpr2
; VI: [[COPY1:%[0-9]+]]:_(p1) = COPY $vgpr3_vgpr4 ; VI: [[COPY1:%[0-9]+]]:_(p1) = COPY $vgpr3_vgpr4
; VI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY]](s96) ; VI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY]](s96)
; VI: G_STORE [[BITCAST]](<3 x s32>), [[COPY1]](p1) :: (store (s96), align 16, addrspace 1) ; VI: G_STORE [[BITCAST]](<3 x s32>), [[COPY1]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
%0:_(s96) = COPY $vgpr0_vgpr1_vgpr2 %0:_(s96) = COPY $vgpr0_vgpr1_vgpr2
%1:_(p1) = COPY $vgpr3_vgpr4 %1:_(p1) = COPY $vgpr3_vgpr4
G_STORE %0, %1 :: (store (s96), addrspace 1, align 16) G_STORE %0, %1 :: (store (s96), addrspace 1, align 16)
... ...
--- ---
name: test_store_global_i128 name: test_store_global_i128
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5 liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; SI-LABEL: name: test_store_global_i128 ; SI-LABEL: name: test_store_global_i128
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), addrspace 1) ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
; VI-LABEL: name: test_store_global_i128 ; VI-LABEL: name: test_store_global_i128
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), addrspace 1) ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (s128), addrspace 1) G_STORE %1, %0 :: (store (s128), addrspace 1)
... ...
--- ---
name: test_store_global_v2s64 name: test_store_global_v2s64
body: | body: |
▲ Show 20 Lines • Show All 62 Lines • ▼ Show 20 Lines bb.0:
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV1]](s32) ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV1]](s32)
; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C2]] ; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C2]]
; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32) ; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; SI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (<2 x s8>), addrspace 1) ; SI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
; VI-LABEL: name: test_store_global_v2s8_align2 ; VI-LABEL: name: test_store_global_v2s8_align2
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[DEF:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF ; VI: [[DEF:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<2 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<2 x s32>)
; VI: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; VI: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]] ; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]]
; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C]] ; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C]]
; VI: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 ; VI: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C1]](s16) ; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C1]](s16)
; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; VI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (<2 x s8>), addrspace 1) ; VI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s8>) = G_IMPLICIT_DEF %1:_(<2 x s8>) = G_IMPLICIT_DEF
G_STORE %1, %0 :: (store (<2 x s8>), addrspace 1, align 2) G_STORE %1, %0 :: (store (<2 x s8>), addrspace 1, align 2)
... ...
--- ---
name: test_store_global_v2s8_align4 name: test_store_global_v2s8_align4
Show All 11 Lines bb.0:
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV1]](s32) ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV1]](s32)
; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C2]] ; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C2]]
; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32) ; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; SI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (<2 x s8>), align 4, addrspace 1) ; SI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (s16), align 4, addrspace 1)
; VI-LABEL: name: test_store_global_v2s8_align4 ; VI-LABEL: name: test_store_global_v2s8_align4
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[DEF:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF ; VI: [[DEF:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<2 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<2 x s32>)
; VI: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 ; VI: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]] ; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]]
; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C]] ; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C]]
; VI: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 ; VI: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C1]](s16) ; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C1]](s16)
; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; VI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (<2 x s8>), align 4, addrspace 1) ; VI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (s16), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s8>) = G_IMPLICIT_DEF %1:_(<2 x s8>) = G_IMPLICIT_DEF
G_STORE %1, %0 :: (store (<2 x s8>), addrspace 1, align 4) G_STORE %1, %0 :: (store (<2 x s8>), addrspace 1, align 4)
... ...
--- ---
name: test_store_global_v3s8_align1 name: test_store_global_v3s8_align1
▲ Show 20 Lines • Show All 335 Lines • ▼ Show 20 Lines bb.0:
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C2]](s32) ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C2]](s32)
; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]] ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
; SI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[UV3]](s32) ; SI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[UV3]](s32)
; SI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C]] ; SI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C]]
; SI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; SI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C3]](s32) ; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C3]](s32)
; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]] ; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
; SI: G_STORE [[OR2]](s32), [[COPY]](p1) :: (store (<4 x s8>), addrspace 1) ; SI: G_STORE [[OR2]](s32), [[COPY]](p1) :: (store (s32), addrspace 1)
; VI-LABEL: name: test_store_global_v4s8_align4 ; VI-LABEL: name: test_store_global_v4s8_align4
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<4 x s32>) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<4 x s32>)
; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32) ; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
; VI: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C]] ; VI: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C]]
; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[UV1]](s32) ; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[UV1]](s32)
; VI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C]] ; VI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C]]
; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; VI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32) ; VI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
; VI: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]] ; VI: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[UV2]](s32) ; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[UV2]](s32)
; VI: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C]] ; VI: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C]]
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C2]](s32) ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C2]](s32)
; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]] ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
; VI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[UV3]](s32) ; VI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[UV3]](s32)
; VI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C]] ; VI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C]]
; VI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C3]](s32) ; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C3]](s32)
; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]] ; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
; VI: G_STORE [[OR2]](s32), [[COPY]](p1) :: (store (<4 x s8>), addrspace 1) ; VI: G_STORE [[OR2]](s32), [[COPY]](p1) :: (store (s32), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %1:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
%2:_(<4 x s8>) = G_TRUNC %1 %2:_(<4 x s8>) = G_TRUNC %1
G_STORE %2, %0 :: (store (<4 x s8>), addrspace 1, align 4) G_STORE %2, %0 :: (store (<4 x s8>), addrspace 1, align 4)
... ...
--- ---
name: test_truncstore_global_v2s8_to_1_align1 name: test_truncstore_global_v2s8_to_1_align1
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5 liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; SI-LABEL: name: test_truncstore_global_v2s8_to_1_align1 ; SI-LABEL: name: test_truncstore_global_v2s8_to_1_align1
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[DEF:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF ; SI: [[DEF:%[0-9]+]]:_(<2 x s8>) = G_IMPLICIT_DEF
; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<2 x s32>) ; SI: G_STORE [[DEF]](<2 x s8>), [[COPY]](p1) :: (store (<2 x s4>), addrspace 1)
; SI: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
; SI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]]
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV1]](s32)
; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C2]]
; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; SI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (<2 x s4>), addrspace 1)
; VI-LABEL: name: test_truncstore_global_v2s8_to_1_align1 ; VI-LABEL: name: test_truncstore_global_v2s8_to_1_align1
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[DEF:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF ; VI: [[DEF:%[0-9]+]]:_(<2 x s8>) = G_IMPLICIT_DEF
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<2 x s32>) ; VI: G_STORE [[DEF]](<2 x s8>), [[COPY]](p1) :: (store (<2 x s4>), addrspace 1)
; VI: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]]
; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C]]
; VI: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C1]](s16)
; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16)
; VI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (<2 x s4>), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s8>) = G_IMPLICIT_DEF %1:_(<2 x s8>) = G_IMPLICIT_DEF
G_STORE %1, %0 :: (store (<2 x s4>), addrspace 1, align 1) G_STORE %1, %0 :: (store (<2 x s4>), addrspace 1, align 1)
... ...
--- ---
name: test_truncstore_global_v3s8_to_1_align1 name: test_truncstore_global_v3s8_to_1_align1
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4 liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
; SI-LABEL: name: test_truncstore_global_v3s8_to_1_align1 ; SI-LABEL: name: test_truncstore_global_v3s8_to_1_align1
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4 ; SI: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>) ; SI: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[COPY1]](<3 x s32>)
; SI: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF ; SI: G_STORE [[TRUNC]](<3 x s8>), [[COPY]](p1) :: (store (<3 x s2>), addrspace 1)
; SI: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
; SI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]]
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[UV1]](s32)
; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C2]]
; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY2]](s32)
; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
; SI: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C]]
; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[DEF]](s32)
; SI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C2]]
; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C1]](s32)
; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[SHL1]](s32)
; SI: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[TRUNC3]]
; SI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
; SI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
; SI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; SI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[OR2]](s32)
; SI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 63
; SI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C4]]
; SI: G_STORE [[AND4]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
; VI-LABEL: name: test_truncstore_global_v3s8_to_1_align1 ; VI-LABEL: name: test_truncstore_global_v3s8_to_1_align1
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4 ; VI: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>) ; VI: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[COPY1]](<3 x s32>)
; VI: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF ; VI: G_STORE [[TRUNC]](<3 x s8>), [[COPY]](p1) :: (store (<3 x s2>), addrspace 1)
; VI: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]]
; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C]]
; VI: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C1]](s16)
; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
; VI: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C]]
; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
; VI: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C]]
; VI: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C1]](s16)
; VI: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
; VI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
; VI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C2]](s32)
; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[OR2]](s32)
; VI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 63
; VI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C3]]
; VI: G_STORE [[AND4]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4 %1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
%2:_(<3 x s8>) = G_TRUNC %1 %2:_(<3 x s8>) = G_TRUNC %1
G_STORE %2, %0 :: (store (<3 x s2>), addrspace 1, align 1) G_STORE %2, %0 :: (store (<3 x s2>), addrspace 1, align 1)
... ...
--- ---
name: test_truncstore_global_v3s8_to_2_align2 name: test_truncstore_global_v3s8_to_2_align2
body: | body: |
bb.0: bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4 liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
; SI-LABEL: name: test_truncstore_global_v3s8_to_2_align2 ; SI-LABEL: name: test_truncstore_global_v3s8_to_2_align2
; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; SI: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4 ; SI: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>) ; SI: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[COPY1]](<3 x s32>)
; SI: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF ; SI: G_STORE [[TRUNC]](<3 x s8>), [[COPY]](p1) :: (store (<3 x s4>), addrspace 1)
; SI: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
; SI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]]
; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[UV1]](s32)
; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C2]]
; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY2]](s32)
; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32)
; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]]
; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
; SI: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C]]
; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[DEF]](s32)
; SI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C2]]
; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C1]](s32)
; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[SHL1]](s32)
; SI: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[TRUNC3]]
; SI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
; SI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
; SI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; SI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[OR2]](s32)
; SI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 4095
; SI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C4]]
; SI: G_STORE [[AND4]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
; VI-LABEL: name: test_truncstore_global_v3s8_to_2_align2 ; VI-LABEL: name: test_truncstore_global_v3s8_to_2_align2
; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; VI: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4 ; VI: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>) ; VI: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[COPY1]](<3 x s32>)
; VI: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF ; VI: G_STORE [[TRUNC]](<3 x s8>), [[COPY]](p1) :: (store (<3 x s4>), addrspace 1)
; VI: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]]
; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C]]
; VI: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C1]](s16)
; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
; VI: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C]]
; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
; VI: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C]]
; VI: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C1]](s16)
; VI: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
; VI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
; VI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C2]](s32)
; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[OR2]](s32)
; VI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4095
; VI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C3]]
; VI: G_STORE [[AND4]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1 %0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4 %1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
%2:_(<3 x s8>) = G_TRUNC %1 %2:_(<3 x s8>) = G_TRUNC %1
G_STORE %2, %0 :: (store (<3 x s4>), addrspace 1, align 2) G_STORE %2, %0 :: (store (<3 x s4>), addrspace 1, align 2)
... ...