This is an archive of the discontinued LLVM Phabricator instance.

Differential D143648

[AMDGPU][GlobalISel] Fix selection of image sample g16 instructions
ClosedPublic

Authored by mbrkusanin on Feb 9 2023, 7:04 AM.

Details

Reviewers
foad
arsenm
Summary

Pre-GFX10 A16 modifier would imply G16. From GFX10 and onward there are
separate instructions for 16bit gradients. This fixes the condition for
selecting G16 opcodes. Also stop adding G16 flag to instructions that do not
use gradients for GFX10 onward.

Diff Detail

Unit TestsFailed

TimeTest
60,080 msx64 debian > libFuzzer.libFuzzer::fuzzer-leak.test
60,130 msx64 debian > libFuzzer.libFuzzer::minimize_crash.test

Event Timeline

mbrkusanin created this revision.Feb 9 2023, 7:04 AM
Herald added a project: Restricted Project. · View Herald TranscriptFeb 9 2023, 7:04 AM
Herald added subscribers: kosarev, StephenFan, wenlei and 7 others. · View Herald Transcript
mbrkusanin requested review of this revision.Feb 9 2023, 7:04 AM
Herald added a subscriber: wdng. · View Herald TranscriptFeb 9 2023, 7:04 AM
arsenm accepted this revision.Feb 9 2023, 7:19 AM
This revision is now accepted and ready to land.Feb 9 2023, 7:19 AM
mbrkusanin closed this revision.Feb 9 2023, 7:42 AM

rG43924cbd29a32dc391f36850405d6c26384acaa1

foad added inline comments.Feb 9 2023, 8:18 AM
llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
4905

This simplifies to GradTy == S16 && (!ST.hasG16() || BaseOpcode->Gradients) - but perhaps your way is more readable.

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.g16.a16.dim.ll
860

How would these new tests fail without your patch? Would they fail machine verification?

mbrkusanin added inline comments.Feb 9 2023, 8:22 AM
llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.g16.a16.dim.ll
860

GlobalISel was selecting non g16 variant without a crash or fail. There were no test cases with a16 and g16.

Harbormaster completed remote builds in B212801: Diff 496115.Feb 9 2023, 9:25 AM

Revision Contents

Diff 496115

llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp

Show First 20 Lines • Show All 1,818 Lines • ▼ Show 20 Lines if (BaseOpcode->Store) {
NumVDataDwords = DMaskLanes; NumVDataDwords = DMaskLanes;
if (IsD16 && !STI.hasUnpackedD16VMem()) if (IsD16 && !STI.hasUnpackedD16VMem())
NumVDataDwords = (DMaskLanes + 1) / 2; NumVDataDwords = (DMaskLanes + 1) / 2;
} }
} }
// Set G16 opcode // Set G16 opcode
if (IsG16 && !IsA16) { if (Subtarget->hasG16() && IsG16) {
const AMDGPU::MIMGG16MappingInfo *G16MappingInfo = const AMDGPU::MIMGG16MappingInfo *G16MappingInfo =
AMDGPU::getMIMGG16MappingInfo(Intr->BaseOpcode); AMDGPU::getMIMGG16MappingInfo(Intr->BaseOpcode);
assert(G16MappingInfo); assert(G16MappingInfo);
IntrOpcode = G16MappingInfo->G16; // set opcode to variant with _g16 IntrOpcode = G16MappingInfo->G16; // set opcode to variant with _g16
} }
// TODO: Check this in verifier. // TODO: Check this in verifier.
assert((!IsTexFail || DMaskLanes >= 1) && "should have legalized this"); assert((!IsTexFail || DMaskLanes >= 1) && "should have legalized this");
▲ Show 20 Lines • Show All 3,251 Lines • Show Last 20 Lines

llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp

Show First 20 Lines • Show All 4,895 Lines • ▼ Show 20 Linesbool AMDGPULegalizerInfo::legalizeImageIntrinsic(
Register VData = MI.getOperand(NumDefs == 0 ? 1 : 0).getReg(); Register VData = MI.getOperand(NumDefs == 0 ? 1 : 0).getReg();
LLT Ty = MRI->getType(VData); LLT Ty = MRI->getType(VData);
// Check for 16 bit addresses and pack if true. // Check for 16 bit addresses and pack if true.
LLT GradTy = LLT GradTy =
MRI->getType(MI.getOperand(ArgOffset + Intr->GradientStart).getReg()); MRI->getType(MI.getOperand(ArgOffset + Intr->GradientStart).getReg());
LLT AddrTy = LLT AddrTy =
MRI->getType(MI.getOperand(ArgOffset + Intr->CoordStart).getReg()); MRI->getType(MI.getOperand(ArgOffset + Intr->CoordStart).getReg());
const bool IsG16 = GradTy == S16; const bool IsG16 =
ST.hasG16() ? (BaseOpcode->Gradients && GradTy == S16) : GradTy == S16;
foadUnsubmitted
Not Done
ReplyInline Actions

This simplifies to GradTy == S16 && (!ST.hasG16() || BaseOpcode->Gradients) - but perhaps your way is more readable.

foad: This simplifies to `GradTy == S16 && (!ST.hasG16() || BaseOpcode->Gradients)` - but perhaps…
const bool IsA16 = AddrTy == S16; const bool IsA16 = AddrTy == S16;
const bool IsD16 = Ty.getScalarType() == S16; const bool IsD16 = Ty.getScalarType() == S16;
int DMaskLanes = 0; int DMaskLanes = 0;
if (!BaseOpcode->Atomic) { if (!BaseOpcode->Atomic) {
DMask = MI.getOperand(ArgOffset + Intr->DMaskIndex).getImm(); DMask = MI.getOperand(ArgOffset + Intr->DMaskIndex).getImm();
if (BaseOpcode->Gather4) { if (BaseOpcode->Gather4) {
DMaskLanes = 4; DMaskLanes = 4;
▲ Show 20 Lines • Show All 900 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-llvm.amdgcn.image.atomic.dim.a16.ll

Show All 33 Linesdefine amdgpu_ps float @atomic_swap_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6 ; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7 ; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.swap.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.swap.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.swap.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.swap.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
Show All 28 Linesdefine amdgpu_ps float @atomic_add_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6 ; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7 ; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.add.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.add.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
Show All 28 Linesdefine amdgpu_ps float @atomic_sub_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6 ; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7 ; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.sub.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.sub.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.sub.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.sub.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
Show All 28 Linesdefine amdgpu_ps float @atomic_smin_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6 ; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7 ; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.smin.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.smin.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.smin.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.smin.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
Show All 29 Linesdefine amdgpu_ps float @atomic_umin_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6 ; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7 ; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.umin.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.umin.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.umin.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.umin.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
Show All 28 Linesdefine amdgpu_ps float @atomic_smax_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6 ; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7 ; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.smax.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.smax.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.smax.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.smax.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
Show All 28 Linesdefine amdgpu_ps float @atomic_umax_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6 ; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7 ; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.umax.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.umax.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.umax.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.umax.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
Show All 28 Linesdefine amdgpu_ps float @atomic_and_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6 ; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7 ; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.and.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.and.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.and.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.and.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
Show All 28 Linesdefine amdgpu_ps float @atomic_or_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6 ; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7 ; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.or.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.or.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.or.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.or.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
Show All 28 Linesdefine amdgpu_ps float @atomic_xor_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6 ; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7 ; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.xor.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.xor.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.xor.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.xor.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
Show All 28 Linesdefine amdgpu_ps float @atomic_inc_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6 ; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7 ; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.inc.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.inc.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.inc.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.inc.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
Show All 28 Linesdefine amdgpu_ps float @atomic_dec_1d(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6 ; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7 ; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.dec.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.dec.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.dec.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.dec.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
Show All 32 Linesdefine amdgpu_ps float @atomic_cmpswap_1d(<8 x i32> inreg %rsrc, i32 %cmp, i32 %swap, i16 %s) {
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.cmpswap.1d), [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.cmpswap.1d), [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.cmpswap.1d.i32.i16(i32 %cmp, i32 %swap, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.cmpswap.1d.i32.i16(i32 %cmp, i32 %swap, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
Show All 34 Linesdefine amdgpu_ps float @atomic_add_2d(<8 x i32> inreg %rsrc, i32 %data, i16 %s, i16 %t) {
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32)
; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.2d), [[COPY8]](s32), [[BUILD_VECTOR1]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.2d), [[COPY8]](s32), [[BUILD_VECTOR1]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.add.2d.i32.i16(i32 %data, i16 %s, i16 %t, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.add.2d.i32.i16(i32 %data, i16 %s, i16 %t, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
▲ Show 20 Lines • Show All 44 Lines • ▼ Show 20 Linesdefine amdgpu_ps float @atomic_add_3d(<8 x i32> inreg %rsrc, i32 %data, i16 %s, i16 %t, i16 %r) {
; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32)
; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3 ; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY11]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY11]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.3d), [[COPY8]](s32), [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.3d), [[COPY8]](s32), [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.add.3d.i32.i16(i32 %data, i16 %s, i16 %t, i16 %r, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.add.3d.i32.i16(i32 %data, i16 %s, i16 %t, i16 %r, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
▲ Show 20 Lines • Show All 44 Lines • ▼ Show 20 Linesdefine amdgpu_ps float @atomic_add_cube(<8 x i32> inreg %rsrc, i32 %data, i16 %s, i16 %t, i16 %face) {
; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32)
; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3 ; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY11]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY11]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.cube), [[COPY8]](s32), [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.cube), [[COPY8]](s32), [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.add.cube.i32.i16(i32 %data, i16 %s, i16 %t, i16 %face, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.add.cube.i32.i16(i32 %data, i16 %s, i16 %t, i16 %face, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
Show All 34 Linesdefine amdgpu_ps float @atomic_add_1darray(<8 x i32> inreg %rsrc, i32 %data, i16 %s, i16 %slice) {
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32)
; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.1darray), [[COPY8]](s32), [[BUILD_VECTOR1]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.1darray), [[COPY8]](s32), [[BUILD_VECTOR1]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.add.1darray.i32.i16(i32 %data, i16 %s, i16 %slice, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.add.1darray.i32.i16(i32 %data, i16 %s, i16 %slice, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
▲ Show 20 Lines • Show All 44 Lines • ▼ Show 20 Linesdefine amdgpu_ps float @atomic_add_2darray(<8 x i32> inreg %rsrc, i32 %data, i16 %s, i16 %t, i16 %slice) {
; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32)
; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3 ; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY11]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY11]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.2darray), [[COPY8]](s32), [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.2darray), [[COPY8]](s32), [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.add.2darray.i32.i16(i32 %data, i16 %s, i16 %t, i16 %slice, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.add.2darray.i32.i16(i32 %data, i16 %s, i16 %t, i16 %slice, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
▲ Show 20 Lines • Show All 44 Lines • ▼ Show 20 Linesdefine amdgpu_ps float @atomic_add_2dmsaa(<8 x i32> inreg %rsrc, i32 %data, i16 %s, i16 %t, i16 %fragid) {
; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32)
; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3 ; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY11]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY11]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.2dmsaa), [[COPY8]](s32), [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.2dmsaa), [[COPY8]](s32), [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.add.2dmsaa.i32.i16(i32 %data, i16 %s, i16 %t, i16 %fragid, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.add.2dmsaa.i32.i16(i32 %data, i16 %s, i16 %t, i16 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
▲ Show 20 Lines • Show All 46 Lines • ▼ Show 20 Linesdefine amdgpu_ps float @atomic_add_2darraymsaa(<8 x i32> inreg %rsrc, i32 %data, i16 %s, i16 %t, i16 %slice, i16 %fragid) {
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32)
; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3 ; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY11]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY11]](s32)
; GFX10NSA-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr4 ; GFX10NSA-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr4
; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32) ; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.2darraymsaa), [[COPY8]](s32), [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.2darraymsaa), [[COPY8]](s32), [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.add.2darraymsaa.i32.i16(i32 %data, i16 %s, i16 %t, i16 %slice, i16 %fragid, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.add.2darraymsaa.i32.i16(i32 %data, i16 %s, i16 %t, i16 %slice, i16 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
Show All 28 Linesdefine amdgpu_ps float @atomic_add_1d_slc(<8 x i32> inreg %rsrc, i32 %data, i16 %s) {
; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6 ; GFX10NSA-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7 ; GFX10NSA-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY9]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 2, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.add.1d), [[COPY8]](s32), [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 2, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.add.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 2) %v = call i32 @llvm.amdgcn.image.atomic.add.1d.i32.i16(i32 %data, i16 %s, <8 x i32> %rsrc, i32 0, i32 2)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
Show All 38 Linesdefine amdgpu_ps float @atomic_cmpswap_2d(<8 x i32> inreg %rsrc, i32 %cmp, i32 %swap, i16 %s, i16 %t) {
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32)
; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3 ; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY11]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY11]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.cmpswap.2d), [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.cmpswap.2d), [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.cmpswap.2d.i32.i16(i32 %cmp, i32 %swap, i16 %s, i16 %t, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.cmpswap.2d.i32.i16(i32 %cmp, i32 %swap, i16 %s, i16 %t, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
▲ Show 20 Lines • Show All 48 Lines • ▼ Show 20 Linesdefine amdgpu_ps float @atomic_cmpswap_3d(<8 x i32> inreg %rsrc, i32 %cmp, i32 %swap, i16 %s, i16 %t, i16 %r) {
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY11]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY11]](s32)
; GFX10NSA-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr4 ; GFX10NSA-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr4
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.cmpswap.3d), [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.cmpswap.3d), [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.cmpswap.3d.i32.i16(i32 %cmp, i32 %swap, i16 %s, i16 %t, i16 %r, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.cmpswap.3d.i32.i16(i32 %cmp, i32 %swap, i16 %s, i16 %t, i16 %r, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
▲ Show 20 Lines • Show All 50 Lines • ▼ Show 20 Linesdefine amdgpu_ps float @atomic_cmpswap_2darraymsaa(<8 x i32> inreg %rsrc, i32 %cmp, i32 %swap, i16 %s, i16 %t, i16 %slice, i16 %fragid) {
; GFX10NSA-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr4 ; GFX10NSA-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr4
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX10NSA-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr5 ; GFX10NSA-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr5
; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.cmpswap.2darraymsaa), [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (volatile dereferenceable load store (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.atomic.cmpswap.2darraymsaa), [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (volatile dereferenceable load store (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%v = call i32 @llvm.amdgcn.image.atomic.cmpswap.2darraymsaa.i32.i16(i32 %cmp, i32 %swap, i16 %s, i16 %t, i16 %slice, i16 %fragid, <8 x i32> %rsrc, i32 0, i32 0) %v = call i32 @llvm.amdgcn.image.atomic.cmpswap.2darraymsaa.i32.i16(i32 %cmp, i32 %swap, i16 %s, i16 %t, i16 %slice, i16 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
%out = bitcast i32 %v to float %out = bitcast i32 %v to float
ret float %out ret float %out
} }
Show All 29 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-llvm.amdgcn.image.dim.a16.ll

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 41 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_1d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
▲ Show 20 Lines • Show All 50 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_2d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10NSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX10NSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32) ; GFX10NSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32)
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 15, [[BUILD_VECTOR1]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 15, [[BUILD_VECTOR1]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
▲ Show 20 Lines • Show All 65 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.3d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.3d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
▲ Show 20 Lines • Show All 66 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_cube(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.cube), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.cube), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
▲ Show 20 Lines • Show All 52 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_1darray(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10NSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX10NSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32) ; GFX10NSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32)
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1darray), 15, [[BUILD_VECTOR1]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1darray), 15, [[BUILD_VECTOR1]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
▲ Show 20 Lines • Show All 65 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_2darray(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2darray), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2darray), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
▲ Show 20 Lines • Show All 66 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_2dmsaa(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2dmsaa), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2dmsaa), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
▲ Show 20 Lines • Show All 70 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_2darraymsaa(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32) ; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32)
; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2darraymsaa), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2darraymsaa), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
▲ Show 20 Lines • Show All 53 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_mip_1d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10NSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX10NSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32) ; GFX10NSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32)
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.mip.1d), 15, [[BUILD_VECTOR1]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.mip.1d), 15, [[BUILD_VECTOR1]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
▲ Show 20 Lines • Show All 65 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_mip_2d(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.mip.2d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.mip.2d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
▲ Show 20 Lines • Show All 70 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_mip_3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32) ; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32)
; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.mip.3d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.mip.3d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
▲ Show 20 Lines • Show All 71 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_mip_cube(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32) ; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32)
; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.mip.cube), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.mip.cube), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
▲ Show 20 Lines • Show All 67 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_mip_1darray(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.mip.1darray), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.mip.1darray), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
▲ Show 20 Lines • Show All 70 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_mip_2darray(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32) ; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32)
; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.mip.2darray), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.mip.2darray), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
▲ Show 20 Lines • Show All 48 Lines • ▼ Show 20 Linesdefine amdgpu_ps void @store_1d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3 ; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX10NSA-NEXT: [[COPY12:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr4 ; GFX10NSA-NEXT: [[COPY12:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr4
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.1d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable store (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.1d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable store (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
call void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) call void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
ret void ret void
} }
define amdgpu_ps void @store_2d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords) { define amdgpu_ps void @store_2d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords) {
▲ Show 20 Lines • Show All 50 Lines • ▼ Show 20 Linesdefine amdgpu_ps void @store_2d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>)
; GFX10NSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10NSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX10NSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32) ; GFX10NSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32)
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable store (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable store (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
%t = extractelement <2 x i16> %coords, i32 1 %t = extractelement <2 x i16> %coords, i32 1
call void @llvm.amdgcn.image.store.2d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, <8 x i32> %rsrc, i32 0, i32 0) call void @llvm.amdgcn.image.store.2d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, <8 x i32> %rsrc, i32 0, i32 0)
ret void ret void
} }
▲ Show 20 Lines • Show All 65 Lines • ▼ Show 20 Linesdefine amdgpu_ps void @store_3d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.3d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable store (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.3d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable store (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
%t = extractelement <2 x i16> %coords_lo, i32 1 %t = extractelement <2 x i16> %coords_lo, i32 1
%r = extractelement <2 x i16> %coords_hi, i32 0 %r = extractelement <2 x i16> %coords_hi, i32 0
call void @llvm.amdgcn.image.store.3d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %r, <8 x i32> %rsrc, i32 0, i32 0) call void @llvm.amdgcn.image.store.3d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %r, <8 x i32> %rsrc, i32 0, i32 0)
ret void ret void
} }
▲ Show 20 Lines • Show All 66 Lines • ▼ Show 20 Linesdefine amdgpu_ps void @store_cube(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.cube), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable store (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.cube), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable store (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
%t = extractelement <2 x i16> %coords_lo, i32 1 %t = extractelement <2 x i16> %coords_lo, i32 1
%slice = extractelement <2 x i16> %coords_hi, i32 0 %slice = extractelement <2 x i16> %coords_hi, i32 0
call void @llvm.amdgcn.image.store.cube.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %slice, <8 x i32> %rsrc, i32 0, i32 0) call void @llvm.amdgcn.image.store.cube.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %slice, <8 x i32> %rsrc, i32 0, i32 0)
ret void ret void
} }
▲ Show 20 Lines • Show All 52 Lines • ▼ Show 20 Linesdefine amdgpu_ps void @store_1darray(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>)
; GFX10NSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10NSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX10NSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32) ; GFX10NSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32)
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.1darray), [[BUILD_VECTOR1]](<4 x s32>), 15, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable store (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.1darray), [[BUILD_VECTOR1]](<4 x s32>), 15, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable store (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
%slice = extractelement <2 x i16> %coords, i32 1 %slice = extractelement <2 x i16> %coords, i32 1
call void @llvm.amdgcn.image.store.1darray.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %slice, <8 x i32> %rsrc, i32 0, i32 0) call void @llvm.amdgcn.image.store.1darray.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %slice, <8 x i32> %rsrc, i32 0, i32 0)
ret void ret void
} }
▲ Show 20 Lines • Show All 65 Lines • ▼ Show 20 Linesdefine amdgpu_ps void @store_2darray(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2darray), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable store (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2darray), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable store (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
%t = extractelement <2 x i16> %coords_lo, i32 1 %t = extractelement <2 x i16> %coords_lo, i32 1
%slice = extractelement <2 x i16> %coords_hi, i32 0 %slice = extractelement <2 x i16> %coords_hi, i32 0
call void @llvm.amdgcn.image.store.2darray.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %slice, <8 x i32> %rsrc, i32 0, i32 0) call void @llvm.amdgcn.image.store.2darray.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %slice, <8 x i32> %rsrc, i32 0, i32 0)
ret void ret void
} }
▲ Show 20 Lines • Show All 66 Lines • ▼ Show 20 Linesdefine amdgpu_ps void @store_2dmsaa(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2dmsaa), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable store (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2dmsaa), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable store (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
%t = extractelement <2 x i16> %coords_lo, i32 1 %t = extractelement <2 x i16> %coords_lo, i32 1
%fragid = extractelement <2 x i16> %coords_hi, i32 0 %fragid = extractelement <2 x i16> %coords_hi, i32 0
call void @llvm.amdgcn.image.store.2dmsaa.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %fragid, <8 x i32> %rsrc, i32 0, i32 0) call void @llvm.amdgcn.image.store.2dmsaa.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
ret void ret void
} }
▲ Show 20 Lines • Show All 70 Lines • ▼ Show 20 Linesdefine amdgpu_ps void @store_2darraymsaa(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32) ; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32)
; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2darraymsaa), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable store (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2darraymsaa), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable store (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
%t = extractelement <2 x i16> %coords_lo, i32 1 %t = extractelement <2 x i16> %coords_lo, i32 1
%slice = extractelement <2 x i16> %coords_hi, i32 0 %slice = extractelement <2 x i16> %coords_hi, i32 0
%fragid = extractelement <2 x i16> %coords_hi, i32 1 %fragid = extractelement <2 x i16> %coords_hi, i32 1
call void @llvm.amdgcn.image.store.2darraymsaa.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %slice, i16 %fragid, <8 x i32> %rsrc, i32 0, i32 0) call void @llvm.amdgcn.image.store.2darraymsaa.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %slice, i16 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
ret void ret void
▲ Show 20 Lines • Show All 53 Lines • ▼ Show 20 Linesdefine amdgpu_ps void @store_mip_1d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>)
; GFX10NSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10NSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX10NSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32) ; GFX10NSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32)
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.mip.1d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable store (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.mip.1d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable store (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
%mip = extractelement <2 x i16> %coords, i32 1 %mip = extractelement <2 x i16> %coords, i32 1
call void @llvm.amdgcn.image.store.mip.1d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0) call void @llvm.amdgcn.image.store.mip.1d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0)
ret void ret void
} }
▲ Show 20 Lines • Show All 65 Lines • ▼ Show 20 Linesdefine amdgpu_ps void @store_mip_2d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.mip.2d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable store (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.mip.2d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable store (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
%t = extractelement <2 x i16> %coords_lo, i32 1 %t = extractelement <2 x i16> %coords_lo, i32 1
%mip = extractelement <2 x i16> %coords_hi, i32 0 %mip = extractelement <2 x i16> %coords_hi, i32 0
call void @llvm.amdgcn.image.store.mip.2d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0) call void @llvm.amdgcn.image.store.mip.2d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0)
ret void ret void
} }
▲ Show 20 Lines • Show All 70 Lines • ▼ Show 20 Linesdefine amdgpu_ps void @store_mip_3d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32) ; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32)
; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.mip.3d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable store (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.mip.3d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable store (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
%t = extractelement <2 x i16> %coords_lo, i32 1 %t = extractelement <2 x i16> %coords_lo, i32 1
%r = extractelement <2 x i16> %coords_hi, i32 0 %r = extractelement <2 x i16> %coords_hi, i32 0
%mip = extractelement <2 x i16> %coords_hi, i32 1 %mip = extractelement <2 x i16> %coords_hi, i32 1
call void @llvm.amdgcn.image.store.mip.3d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %r, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0) call void @llvm.amdgcn.image.store.mip.3d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %r, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0)
ret void ret void
▲ Show 20 Lines • Show All 71 Lines • ▼ Show 20 Linesdefine amdgpu_ps void @store_mip_cube(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32) ; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32)
; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.mip.cube), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable store (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.mip.cube), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable store (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
%t = extractelement <2 x i16> %coords_lo, i32 1 %t = extractelement <2 x i16> %coords_lo, i32 1
%slice = extractelement <2 x i16> %coords_hi, i32 0 %slice = extractelement <2 x i16> %coords_hi, i32 0
%mip = extractelement <2 x i16> %coords_hi, i32 1 %mip = extractelement <2 x i16> %coords_hi, i32 1
call void @llvm.amdgcn.image.store.mip.cube.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %slice, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0) call void @llvm.amdgcn.image.store.mip.cube.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %slice, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0)
ret void ret void
▲ Show 20 Lines • Show All 67 Lines • ▼ Show 20 Linesdefine amdgpu_ps void @store_mip_1darray(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.mip.1darray), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable store (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.mip.1darray), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable store (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
%slice = extractelement <2 x i16> %coords_lo, i32 1 %slice = extractelement <2 x i16> %coords_lo, i32 1
%mip = extractelement <2 x i16> %coords_hi, i32 0 %mip = extractelement <2 x i16> %coords_hi, i32 0
call void @llvm.amdgcn.image.store.mip.1darray.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %slice, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0) call void @llvm.amdgcn.image.store.mip.1darray.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %slice, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0)
ret void ret void
} }
▲ Show 20 Lines • Show All 70 Lines • ▼ Show 20 Linesdefine amdgpu_ps void @store_mip_2darray(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY13]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32) ; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32)
; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.mip.2darray), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable store (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.mip.2darray), [[BUILD_VECTOR1]](<4 x s32>), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable store (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords_lo, i32 0 %s = extractelement <2 x i16> %coords_lo, i32 0
%t = extractelement <2 x i16> %coords_lo, i32 1 %t = extractelement <2 x i16> %coords_lo, i32 1
%slice = extractelement <2 x i16> %coords_hi, i32 0 %slice = extractelement <2 x i16> %coords_hi, i32 0
%mip = extractelement <2 x i16> %coords_hi, i32 1 %mip = extractelement <2 x i16> %coords_hi, i32 1
call void @llvm.amdgcn.image.store.mip.2darray.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %slice, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0) call void @llvm.amdgcn.image.store.mip.2darray.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, i16 %t, i16 %slice, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0)
ret void ret void
Show All 38 Linesdefine amdgpu_ps <4 x float> @getresinfo_1d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.getresinfo.1d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.getresinfo.1d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%mip = extractelement <2 x i16> %coords, i32 0 %mip = extractelement <2 x i16> %coords, i32 0
Show All 40 Linesdefine amdgpu_ps <4 x float> @getresinfo_2d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.getresinfo.2d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.getresinfo.2d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%mip = extractelement <2 x i16> %coords, i32 0 %mip = extractelement <2 x i16> %coords, i32 0
Show All 40 Linesdefine amdgpu_ps <4 x float> @getresinfo_3d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.getresinfo.3d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.getresinfo.3d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%mip = extractelement <2 x i16> %coords, i32 0 %mip = extractelement <2 x i16> %coords, i32 0
Show All 40 Linesdefine amdgpu_ps <4 x float> @getresinfo_cube(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.getresinfo.cube), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.getresinfo.cube), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%mip = extractelement <2 x i16> %coords, i32 0 %mip = extractelement <2 x i16> %coords, i32 0
Show All 40 Linesdefine amdgpu_ps <4 x float> @getresinfo_1darray(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.getresinfo.1darray), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.getresinfo.1darray), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%mip = extractelement <2 x i16> %coords, i32 0 %mip = extractelement <2 x i16> %coords, i32 0
Show All 40 Linesdefine amdgpu_ps <4 x float> @getresinfo_2darray(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.getresinfo.2darray), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.getresinfo.2darray), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%mip = extractelement <2 x i16> %coords, i32 0 %mip = extractelement <2 x i16> %coords, i32 0
Show All 40 Linesdefine amdgpu_ps <4 x float> @getresinfo_2dmsaa(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.getresinfo.2dmsaa), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.getresinfo.2dmsaa), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%mip = extractelement <2 x i16> %coords, i32 0 %mip = extractelement <2 x i16> %coords, i32 0
Show All 40 Linesdefine amdgpu_ps <4 x float> @getresinfo_2darraymsaa(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.getresinfo.2darraymsaa), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.getresinfo.2darraymsaa), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%mip = extractelement <2 x i16> %coords, i32 0 %mip = extractelement <2 x i16> %coords, i32 0
Show All 36 Linesdefine amdgpu_ps float @load_1d_V1(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 8, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable load (s32), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 8, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable load (s32), addrspace 7)
; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
%v = call float @llvm.amdgcn.image.load.1d.f32.i16(i32 8, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) %v = call float @llvm.amdgcn.image.load.1d.f32.i16(i32 8, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
ret float %v ret float %v
} }
Show All 34 Linesdefine amdgpu_ps <2 x float> @load_1d_V2(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<2 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 9, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable load (<2 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<2 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 9, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable load (<2 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<2 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<2 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
%v = call <2 x float> @llvm.amdgcn.image.load.1d.v2f32.i16(i32 9, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) %v = call <2 x float> @llvm.amdgcn.image.load.1d.v2f32.i16(i32 9, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
ret <2 x float> %v ret <2 x float> %v
Show All 35 Linesdefine amdgpu_ps void @store_1d_V1(<8 x i32> inreg %rsrc, float %vdata, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.1d), [[COPY8]](s32), 2, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable store (s32), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.1d), [[COPY8]](s32), 2, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable store (s32), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
call void @llvm.amdgcn.image.store.1d.f32.i16(float %vdata, i32 2, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) call void @llvm.amdgcn.image.store.1d.f32.i16(float %vdata, i32 2, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
ret void ret void
} }
define amdgpu_ps void @store_1d_V2(<8 x i32> inreg %rsrc, <2 x float> %vdata, <2 x i16> %coords) { define amdgpu_ps void @store_1d_V2(<8 x i32> inreg %rsrc, <2 x float> %vdata, <2 x i16> %coords) {
Show All 36 Linesdefine amdgpu_ps void @store_1d_V2(<8 x i32> inreg %rsrc, <2 x float> %vdata, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10NSA-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2 ; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY10]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY10]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.1d), [[BUILD_VECTOR1]](<2 x s32>), 12, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 3 :: (dereferenceable store (<2 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.1d), [[BUILD_VECTOR1]](<2 x s32>), 12, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 1 :: (dereferenceable store (<2 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
call void @llvm.amdgcn.image.store.1d.v2f32.i16(<2 x float> %vdata, i32 12, i16 %s, <8 x i32> %rsrc, i32 0, i32 0) call void @llvm.amdgcn.image.store.1d.v2f32.i16(<2 x float> %vdata, i32 12, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
ret void ret void
} }
define amdgpu_ps <4 x float> @load_1d_glc(<8 x i32> inreg %rsrc, <2 x i16> %coords) { define amdgpu_ps <4 x float> @load_1d_glc(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
Show All 35 Linesdefine amdgpu_ps <4 x float> @load_1d_glc(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 1, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 1, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
Show All 40 Linesdefine amdgpu_ps <4 x float> @load_1d_slc(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 2, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 2, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
Show All 40 Linesdefine amdgpu_ps <4 x float> @load_1d_glc_slc(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8 ; GFX10NSA-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 3, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 3, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
▲ Show 20 Lines • Show All 45 Lines • ▼ Show 20 Linesdefine amdgpu_ps void @store_1d_glc(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3 ; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX10NSA-NEXT: [[COPY12:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr4 ; GFX10NSA-NEXT: [[COPY12:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr4
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.1d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 1, 3 :: (dereferenceable store (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.1d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 1, 1 :: (dereferenceable store (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
call void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, <8 x i32> %rsrc, i32 0, i32 1) call void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, <8 x i32> %rsrc, i32 0, i32 1)
ret void ret void
} }
define amdgpu_ps void @store_1d_slc(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords) { define amdgpu_ps void @store_1d_slc(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords) {
Show All 40 Linesdefine amdgpu_ps void @store_1d_slc(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3 ; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX10NSA-NEXT: [[COPY12:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr4 ; GFX10NSA-NEXT: [[COPY12:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr4
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.1d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 2, 3 :: (dereferenceable store (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.1d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 2, 1 :: (dereferenceable store (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
call void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, <8 x i32> %rsrc, i32 0, i32 2) call void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, <8 x i32> %rsrc, i32 0, i32 2)
ret void ret void
} }
define amdgpu_ps void @store_1d_glc_slc(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords) { define amdgpu_ps void @store_1d_glc_slc(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords) {
Show All 40 Linesdefine amdgpu_ps void @store_1d_glc_slc(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10NSA-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3 ; GFX10NSA-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX10NSA-NEXT: [[COPY12:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr4 ; GFX10NSA-NEXT: [[COPY12:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr4
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY12]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.1d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 3, 3 :: (dereferenceable store (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.1d), [[BUILD_VECTOR1]](<4 x s32>), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 0, 3, 1 :: (dereferenceable store (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: S_ENDPGM 0 ; GFX10NSA-NEXT: S_ENDPGM 0
main_body: main_body:
%s = extractelement <2 x i16> %coords, i32 0 %s = extractelement <2 x i16> %coords, i32 0
call void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, <8 x i32> %rsrc, i32 0, i32 3) call void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, <8 x i32> %rsrc, i32 0, i32 3)
ret void ret void
} }
define amdgpu_ps <4 x float> @getresinfo_dmask0(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords) { define amdgpu_ps <4 x float> @getresinfo_dmask0(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x i16> %coords) {
▲ Show 20 Lines • Show All 67 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_1d_tfe(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9 ; GFX10NSA-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10NSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 ; GFX10NSA-NEXT: [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GFX10NSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<5 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 1, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<5 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), 1, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<5 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<5 x s32>)
; GFX10NSA-NEXT: G_STORE [[UV4]](s32), [[DEF]](p1) :: (store (s32) into `ptr addrspace(1) undef`, addrspace 1) ; GFX10NSA-NEXT: G_STORE [[UV4]](s32), [[DEF]](p1) :: (store (s32) into `ptr addrspace(1) undef`, addrspace 1)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
▲ Show 20 Lines • Show All 57 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_2d_tfe(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) ; GFX10NSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX10NSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY8]](<2 x s16>)
; GFX10NSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10NSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX10NSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32) ; GFX10NSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32)
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<5 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 15, [[BUILD_VECTOR1]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 1, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<5 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 15, [[BUILD_VECTOR1]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 1, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<5 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<5 x s32>)
; GFX10NSA-NEXT: G_STORE [[UV4]](s32), [[DEF]](p1) :: (store (s32) into `ptr addrspace(1) undef`, addrspace 1) ; GFX10NSA-NEXT: G_STORE [[UV4]](s32), [[DEF]](p1) :: (store (s32) into `ptr addrspace(1) undef`, addrspace 1)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
▲ Show 20 Lines • Show All 72 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_3d_tfe(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) ; GFX10NSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[DEF1:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10NSA-NEXT: [[DEF1:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF1]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<5 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.3d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 1, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<5 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.3d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 1, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<5 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<5 x s32>)
; GFX10NSA-NEXT: G_STORE [[UV4]](s32), [[DEF]](p1) :: (store (s32) into `ptr addrspace(1) undef`, addrspace 1) ; GFX10NSA-NEXT: G_STORE [[UV4]](s32), [[DEF]](p1) :: (store (s32) into `ptr addrspace(1) undef`, addrspace 1)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
▲ Show 20 Lines • Show All 77 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @load_2darraymsaa_tfe(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10NSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) ; GFX10NSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>) ; GFX10NSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY9]](<2 x s16>)
; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32) ; GFX10NSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32)
; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) ; GFX10NSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16) ; GFX10NSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10NSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<5 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2darraymsaa), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 1, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10NSA-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<5 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2darraymsaa), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), 1, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<5 x s32>) ; GFX10NSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<5 x s32>)
; GFX10NSA-NEXT: G_STORE [[UV4]](s32), [[DEF]](p1) :: (store (s32) into `ptr addrspace(1) undef`, addrspace 1) ; GFX10NSA-NEXT: G_STORE [[UV4]](s32), [[DEF]](p1) :: (store (s32) into `ptr addrspace(1) undef`, addrspace 1)
; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10NSA-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10NSA-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10NSA-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10NSA-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10NSA-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
▲ Show 20 Lines • Show All 61 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-llvm.amdgcn.image.sample.a16.ll

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 45 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %s) {
; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10 ; GFX10-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
; GFX10-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11 ; GFX10-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
; GFX10-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12 ; GFX10-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
; GFX10-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13 ; GFX10-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32) ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.1d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.1d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.1d.v4f32.f16(i32 15, half %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.1d.v4f32.f16(i32 15, half %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 49 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %s, half %t) {
; GFX10-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12 ; GFX10-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
; GFX10-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13 ; GFX10-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32) ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.2d.v4f32.f16(i32 15, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.2d.v4f32.f16(i32 15, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 59 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_3d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %s, half %t, half %r) {
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.3d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.3d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.3d.v4f32.f16(i32 15, half %s, half %t, half %r, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.3d.v4f32.f16(i32 15, half %s, half %t, half %r, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 59 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_cube(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %s, half %t, half %face) {
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cube), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cube), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.cube.v4f32.f16(i32 15, half %s, half %t, half %face, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.cube.v4f32.f16(i32 15, half %s, half %t, half %face, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 49 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_1darray(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %s, half %slice) {
; GFX10-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12 ; GFX10-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
; GFX10-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13 ; GFX10-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32) ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.1darray), 15, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.1darray), 15, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.1darray.v4f32.f16(i32 15, half %s, half %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.1darray.v4f32.f16(i32 15, half %s, half %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 59 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_2darray(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %s, half %t, half %slice) {
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.2darray), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.2darray), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.2darray.v4f32.f16(i32 15, half %s, half %t, half %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.2darray.v4f32.f16(i32 15, half %s, half %t, half %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 53 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_c_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %s) {
; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32) ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32) ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.1d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.1d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.1d.v4f32.f16(i32 15, float %zcompare, half %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.1d.v4f32.f16(i32 15, float %zcompare, half %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 55 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_c_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %s, half %t) {
; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32) ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.2d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.2d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.2d.v4f32.f16(i32 15, float %zcompare, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.2d.v4f32.f16(i32 15, float %zcompare, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 49 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %s, half %clamp) {
; GFX10-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12 ; GFX10-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
; GFX10-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13 ; GFX10-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32) ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cl.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cl.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.cl.1d.v4f32.f16(i32 15, half %s, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.cl.1d.v4f32.f16(i32 15, half %s, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 59 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %s, half %t, half %clamp) {
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cl.2d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cl.2d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.cl.2d.v4f32.f16(i32 15, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.cl.2d.v4f32.f16(i32 15, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 55 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_c_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %s, half %clamp) {
; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32) ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.cl.1d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.cl.1d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.cl.1d.v4f32.f16(i32 15, float %zcompare, half %s, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.cl.1d.v4f32.f16(i32 15, float %zcompare, half %s, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 62 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_c_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %s, half %t, half %clamp) {
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3 ; GFX10-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32) ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.cl.2d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.cl.2d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.cl.2d.v4f32.f16(i32 15, float %zcompare, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.cl.2d.v4f32.f16(i32 15, float %zcompare, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 55 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_b_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %bias, half %s) {
; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.b.1d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.b.1d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.b.1d.v4f32.f16.f16(i32 15, half %bias, half %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.b.1d.v4f32.f16.f16(i32 15, half %bias, half %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 59 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_b_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %bias, half %s, half %t) {
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[TRUNC2]](s16) ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[TRUNC2]](s16)
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.b.2d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.b.2d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.b.2d.v4f32.f16.f16(i32 15, half %bias, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.b.2d.v4f32.f16.f16(i32 15, half %bias, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 58 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_c_b_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %bias, float %zcompare, half %s) {
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY13]](s32) ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY13]](s32)
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.b.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.b.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.b.1d.v4f32.f16.f16(i32 15, half %bias, float %zcompare, half %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.b.1d.v4f32.f16.f16(i32 15, half %bias, float %zcompare, half %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 62 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_c_b_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %bias, float %zcompare, half %s, half %t) {
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3 ; GFX10-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY13]](s32) ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY13]](s32)
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[TRUNC2]](s16) ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[TRUNC2]](s16)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.b.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.b.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.b.2d.v4f32.f16.f16(i32 15, half %bias, float %zcompare, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.b.2d.v4f32.f16.f16(i32 15, half %bias, float %zcompare, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 59 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_b_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %bias, half %s, half %clamp) {
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[TRUNC2]](s16) ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[TRUNC2]](s16)
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.b.cl.1d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.b.cl.1d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.b.cl.1d.v4f32.f16.f16(i32 15, half %bias, half %s, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.b.cl.1d.v4f32.f16.f16(i32 15, half %bias, half %s, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 64 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_b_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %bias, half %s, half %t, half %clamp) {
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3 ; GFX10-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32) ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[TRUNC2]](s16) ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[TRUNC2]](s16)
; GFX10-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC3]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC3]](s16), [[DEF]](s16)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.b.cl.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.b.cl.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.b.cl.2d.v4f32.f16.f16(i32 15, half %bias, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.b.cl.2d.v4f32.f16.f16(i32 15, half %bias, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 62 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_c_b_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %bias, float %zcompare, half %s, half %clamp) {
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3 ; GFX10-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY13]](s32) ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY13]](s32)
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[TRUNC2]](s16) ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[TRUNC2]](s16)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.b.cl.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.b.cl.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.b.cl.1d.v4f32.f16.f16(i32 15, half %bias, float %zcompare, half %s, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.b.cl.1d.v4f32.f16.f16(i32 15, half %bias, float %zcompare, half %s, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 68 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_c_b_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %bias, float %zcompare, half %s, half %t, half %clamp) {
; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
; GFX10-NEXT: [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4 ; GFX10-NEXT: [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32) ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY13]](s32) ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY13]](s32)
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[TRUNC2]](s16) ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[TRUNC2]](s16)
; GFX10-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC3]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC3]](s16), [[DEF]](s16)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.b.cl.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.b.cl.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.b.cl.2d.v4f32.f16.f16(i32 15, half %bias, float %zcompare, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.b.cl.2d.v4f32.f16.f16(i32 15, half %bias, float %zcompare, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 1,549 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_l_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %s, half %lod) {
; GFX10-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12 ; GFX10-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
; GFX10-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13 ; GFX10-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32) ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.l.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.l.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.l.1d.v4f32.f16(i32 15, half %s, half %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.l.1d.v4f32.f16(i32 15, half %s, half %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 59 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_l_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %s, half %t, half %lod) {
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>) ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.l.2d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.l.2d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.l.2d.v4f32.f16(i32 15, half %s, half %t, half %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.l.2d.v4f32.f16(i32 15, half %s, half %t, half %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 55 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_c_l_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %s, half %lod) {
; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32) ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.l.1d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.l.1d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.l.1d.v4f32.f16(i32 15, float %zcompare, half %s, half %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.l.1d.v4f32.f16(i32 15, float %zcompare, half %s, half %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 62 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_c_l_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %s, half %t, half %lod) {
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3 ; GFX10-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32) ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32) ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.l.2d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.l.2d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.l.2d.v4f32.f16(i32 15, float %zcompare, half %s, half %t, half %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.l.2d.v4f32.f16(i32 15, float %zcompare, half %s, half %t, half %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 43 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_lz_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %s) {
; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32) ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10 ; GFX10-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
; GFX10-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11 ; GFX10-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
; GFX10-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12 ; GFX10-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
; GFX10-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13 ; GFX10-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32) ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.lz.1d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.lz.1d), 15, [[TRUNC]](s16), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.lz.1d.v4f32.f16(i32 15, half %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.lz.1d.v4f32.f16(i32 15, half %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 49 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_lz_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %s, half %t) {
; GFX10-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12 ; GFX10-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
; GFX10-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13 ; GFX10-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32) ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.lz.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.lz.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.lz.2d.v4f32.f16(i32 15, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.lz.2d.v4f32.f16(i32 15, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 53 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_c_lz_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %s) {
; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32) ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32) ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.lz.1d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.lz.1d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.lz.1d.v4f32.f16(i32 15, float %zcompare, half %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.lz.1d.v4f32.f16(i32 15, float %zcompare, half %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 55 Lines • ▼ Show 20 Linesdefine amdgpu_ps <4 x float> @sample_c_lz_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %s, half %t) {
; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0 ; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1 ; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32) ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2 ; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32) ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32) ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>) ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.lz.2d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7) ; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.lz.2d), 15, [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 1 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>) ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32) ; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32) ; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32) ; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32) ; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3 ; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body: main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.c.lz.2d.v4f32.f16(i32 15, float %zcompare, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0) %v = call <4 x float> @llvm.amdgcn.image.sample.c.lz.2d.v4f32.f16(i32 15, float %zcompare, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
▲ Show 20 Lines • Show All 258 Lines • Show Last 20 Lines

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-llvm.amdgcn.image.sample.g16.a16.ll

  • This file was added.
; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -stop-after=legalizer -o - %s | FileCheck -check-prefix=GFX10 %s
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -stop-after=legalizer -o - %s | FileCheck -check-prefix=GFX11 %s
define amdgpu_ps <4 x float> @sample_d_1d_g16_a16(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dsdv, half %s) {
; GFX10-LABEL: name: sample_d_1d_g16_a16
; GFX10: bb.1.main_body:
; GFX10-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
; GFX10-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX10-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX10-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
; GFX10-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
; GFX10-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
; GFX10-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
; GFX10-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
; GFX11-LABEL: name: sample_d_1d_g16_a16
; GFX11: bb.1.main_body:
; GFX11-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2
; GFX11-NEXT: {{ $}}
; GFX11-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
; GFX11-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
; GFX11-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
; GFX11-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX11-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX11-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX11-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX11-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
; GFX11-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
; GFX11-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
; GFX11-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX11-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX11-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX11-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX11-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
; GFX11-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
; GFX11-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX11-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX11-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX11-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX11-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX11-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX11-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.1d.v4f32.f16.f16(i32 15, half %dsdh, half %dsdv, half %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v
}
define amdgpu_ps <4 x float> @sample_d_2d_g16_a16(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t) {
; GFX10-LABEL: name: sample_d_2d_g16_a16
; GFX10: bb.1.main_body:
; GFX10-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
; GFX10-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX10-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX10-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
; GFX10-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
; GFX10-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
; GFX10-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
; GFX10-NEXT: [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
; GFX10-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
; GFX10-NEXT: [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
; GFX10-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[COPY17]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
; GFX10-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
; GFX11-LABEL: name: sample_d_2d_g16_a16
; GFX11: bb.1.main_body:
; GFX11-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
; GFX11-NEXT: {{ $}}
; GFX11-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
; GFX11-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
; GFX11-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
; GFX11-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX11-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX11-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX11-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX11-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
; GFX11-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
; GFX11-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
; GFX11-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX11-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX11-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX11-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX11-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
; GFX11-NEXT: [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
; GFX11-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
; GFX11-NEXT: [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
; GFX11-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[COPY17]](s32)
; GFX11-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX11-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
; GFX11-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
; GFX11-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX11-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX11-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX11-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX11-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX11-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f16.f16(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v
}
define amdgpu_ps <4 x float> @sample_d_3d_g16_a16(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %drdh, half %dsdv, half %dtdv, half %drdv, half %s, half %t, half %r) {
; GFX10-LABEL: name: sample_d_3d_g16_a16
; GFX10: bb.1.main_body:
; GFX10-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
; GFX10-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX10-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX10-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX10-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX10-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
; GFX10-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
; GFX10-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
; GFX10-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX10-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX10-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX10-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX10-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
; GFX10-NEXT: [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
; GFX10-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
; GFX10-NEXT: [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
; GFX10-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[COPY17]](s32)
; GFX10-NEXT: [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
; GFX10-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[COPY18]](s32)
; GFX10-NEXT: [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr7
; GFX10-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[COPY19]](s32)
; GFX10-NEXT: [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr8
; GFX10-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[COPY20]](s32)
; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX10-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC3]](s16), [[TRUNC4]](s16)
; GFX10-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC5]](s16), [[DEF]](s16)
; GFX10-NEXT: [[BUILD_VECTOR6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
; GFX10-NEXT: [[BUILD_VECTOR7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC8]](s16), [[DEF]](s16)
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>), [[BUILD_VECTOR6]](<2 x s16>), [[BUILD_VECTOR7]](<2 x s16>)
; GFX10-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.3d), 15, [[CONCAT_VECTORS]](<12 x s16>), $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX10-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX10-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX10-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX10-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX10-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
; GFX11-LABEL: name: sample_d_3d_g16_a16
; GFX11: bb.1.main_body:
; GFX11-NEXT: liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
; GFX11-NEXT: {{ $}}
; GFX11-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
; GFX11-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
; GFX11-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
; GFX11-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
; GFX11-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
; GFX11-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
; GFX11-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
; GFX11-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
; GFX11-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
; GFX11-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
; GFX11-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
; GFX11-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
; GFX11-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
; GFX11-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
; GFX11-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
; GFX11-NEXT: [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
; GFX11-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
; GFX11-NEXT: [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
; GFX11-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[COPY17]](s32)
; GFX11-NEXT: [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
; GFX11-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[COPY18]](s32)
; GFX11-NEXT: [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr7
; GFX11-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[COPY19]](s32)
; GFX11-NEXT: [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr8
; GFX11-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[COPY20]](s32)
; GFX11-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
; GFX11-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
; GFX11-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC3]](s16), [[TRUNC4]](s16)
; GFX11-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC5]](s16), [[DEF]](s16)
; GFX11-NEXT: [[BUILD_VECTOR6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
; GFX11-NEXT: [[BUILD_VECTOR7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC8]](s16), [[DEF]](s16)
; GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>), [[BUILD_VECTOR6]](<2 x s16>), [[BUILD_VECTOR7]](<2 x s16>)
; GFX11-NEXT: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.3d), 15, [[CONCAT_VECTORS]](<12 x s16>), $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 3 :: (dereferenceable load (<4 x s32>), addrspace 7)
; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
; GFX11-NEXT: $vgpr0 = COPY [[UV]](s32)
; GFX11-NEXT: $vgpr1 = COPY [[UV1]](s32)
; GFX11-NEXT: $vgpr2 = COPY [[UV2]](s32)
; GFX11-NEXT: $vgpr3 = COPY [[UV3]](s32)
; GFX11-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.3d.v4f32.f16.f16(i32 15, half %dsdh, half %dtdh, half %drdh, half %dsdv, half %dtdv, half %drdv, half %s, half %t, half %r, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v
}
declare <4 x float> @llvm.amdgcn.image.sample.d.1d.v4f32.f16.f16(i32, half, half, half, <8 x i32>, <4 x i32>, i1, i32, i32)
declare <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f16.f16(i32, half, half, half, half, half, half, <8 x i32>, <4 x i32>, i1, i32, i32)
declare <4 x float> @llvm.amdgcn.image.sample.d.3d.v4f32.f16.f16(i32, half, half, half, half, half, half, half, half, half, <8 x i32>, <4 x i32>, i1, i32, i32)

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.g16.a16.dim.ll

Show First 20 Lines • Show All 851 Lines • ▼ Show 20 Lines
declare <4 x float> @llvm.amdgcn.image.sample.cd.cl.1d.v4f32.f16.f32(i32, half, half, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1 declare <4 x float> @llvm.amdgcn.image.sample.cd.cl.1d.v4f32.f16.f32(i32, half, half, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
declare <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f16.f32(i32, half, half, half, half, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1 declare <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f16.f32(i32, half, half, half, half, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
declare <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.1d.v4f32.f16.f32(i32, float, half, half, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1 declare <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.1d.v4f32.f16.f32(i32, float, half, half, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
declare <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.2d.v4f32.f16.f32(i32, float, half, half, half, half, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1 declare <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.2d.v4f32.f16.f32(i32, float, half, half, half, half, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
declare float @llvm.amdgcn.image.sample.c.d.o.2darray.f32.f16.f32(i32, i32, float, half, half, half, half, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1 declare float @llvm.amdgcn.image.sample.c.d.o.2darray.f32.f16.f32(i32, i32, float, half, half, half, half, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
declare <2 x float> @llvm.amdgcn.image.sample.c.d.o.2darray.v2f32.f16.f32(i32, i32, float, half, half, half, half, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1 declare <2 x float> @llvm.amdgcn.image.sample.c.d.o.2darray.v2f32.f16.f32(i32, i32, float, half, half, half, half, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
define amdgpu_ps <4 x float> @sample_d_1d_g16_a16(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dsdv, half %s) {
foadUnsubmitted
Not Done
ReplyInline Actions

How would these new tests fail without your patch? Would they fail machine verification?

foad: How would these new tests fail without your patch? Would they fail machine verification?
mbrkusaninAuthorUnsubmitted
Done
ReplyInline Actions

GlobalISel was selecting non g16 variant without a crash or fail. There were no test cases with a16 and g16.

mbrkusanin: GlobalISel was selecting non g16 variant without a crash or fail. There were no test cases with…
; GFX10-LABEL: sample_d_1d_g16_a16:
; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: image_sample_d_g16 v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog
;
; GFX10GISEL-LABEL: sample_d_1d_g16_a16:
; GFX10GISEL: ; %bb.0: ; %main_body
; GFX10GISEL-NEXT: image_sample_d_g16 v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16
; GFX10GISEL-NEXT: s_waitcnt vmcnt(0)
; GFX10GISEL-NEXT: ; return to shader part epilog
main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.1d.v4f32.f16.f16(i32 15, half %dsdh, half %dsdv, half %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v
}
define amdgpu_ps <4 x float> @sample_d_2d_g16_a16(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t) {
; GFX10-LABEL: sample_d_2d_g16_a16:
; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_perm_b32 v4, v5, v4, 0x5040100
; GFX10-NEXT: v_perm_b32 v3, v3, v2, 0x5040100
; GFX10-NEXT: v_perm_b32 v2, v1, v0, 0x5040100
; GFX10-NEXT: image_sample_d_g16 v[0:3], v[2:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog
;
; GFX10GISEL-LABEL: sample_d_2d_g16_a16:
; GFX10GISEL: ; %bb.0: ; %main_body
; GFX10GISEL-NEXT: v_perm_b32 v0, v1, v0, 0x5040100
; GFX10GISEL-NEXT: v_perm_b32 v1, v3, v2, 0x5040100
; GFX10GISEL-NEXT: v_perm_b32 v2, v5, v4, 0x5040100
; GFX10GISEL-NEXT: image_sample_d_g16 v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
; GFX10GISEL-NEXT: s_waitcnt vmcnt(0)
; GFX10GISEL-NEXT: ; return to shader part epilog
main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f16.f16(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v
}
define amdgpu_ps <4 x float> @sample_d_3d_g16_a16(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %drdh, half %dsdv, half %dtdv, half %drdv, half %s, half %t, half %r) {
; GFX10-LABEL: sample_d_3d_g16_a16:
; GFX10: ; %bb.0: ; %main_body
; GFX10-NEXT: v_mov_b32_e32 v12, v8
; GFX10-NEXT: v_mov_b32_e32 v10, v5
; GFX10-NEXT: v_mov_b32_e32 v8, v2
; GFX10-NEXT: v_perm_b32 v11, v7, v6, 0x5040100
; GFX10-NEXT: v_perm_b32 v9, v4, v3, 0x5040100
; GFX10-NEXT: v_perm_b32 v7, v1, v0, 0x5040100
; GFX10-NEXT: image_sample_d_g16 v[0:3], v[7:12], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_3D a16
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: ; return to shader part epilog
;
; GFX10GISEL-LABEL: sample_d_3d_g16_a16:
; GFX10GISEL: ; %bb.0: ; %main_body
; GFX10GISEL-NEXT: v_mov_b32_e32 v9, v3
; GFX10GISEL-NEXT: v_mov_b32_e32 v10, v7
; GFX10GISEL-NEXT: v_mov_b32_e32 v3, v2
; GFX10GISEL-NEXT: v_mov_b32_e32 v7, v8
; GFX10GISEL-NEXT: v_perm_b32 v2, v1, v0, 0x5040100
; GFX10GISEL-NEXT: v_perm_b32 v4, v4, v9, 0x5040100
; GFX10GISEL-NEXT: v_perm_b32 v6, v10, v6, 0x5040100
; GFX10GISEL-NEXT: image_sample_d_g16 v[0:3], v[2:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_3D a16
; GFX10GISEL-NEXT: s_waitcnt vmcnt(0)
; GFX10GISEL-NEXT: ; return to shader part epilog
main_body:
%v = call <4 x float> @llvm.amdgcn.image.sample.d.3d.v4f32.f16.f16(i32 15, half %dsdh, half %dtdh, half %drdh, half %dsdv, half %dtdv, half %drdv, half %s, half %t, half %r, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
ret <4 x float> %v
}
declare <4 x float> @llvm.amdgcn.image.sample.d.1d.v4f32.f16.f16(i32, half, half, half, <8 x i32>, <4 x i32>, i1, i32, i32)
declare <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f16.f16(i32, half, half, half, half, half, half, <8 x i32>, <4 x i32>, i1, i32, i32)
declare <4 x float> @llvm.amdgcn.image.sample.d.3d.v4f32.f16.f16(i32, half, half, half, half, half, half, half, half, half, <8 x i32>, <4 x i32>, i1, i32, i32)
attributes #0 = { nounwind } attributes #0 = { nounwind }
attributes #1 = { nounwind readonly } attributes #1 = { nounwind readonly }
attributes #2 = { nounwind readnone } attributes #2 = { nounwind readnone }