This is an archive of the discontinued LLVM Phabricator instance.

Differential D75187

[AMDGPU][ConstantFolding] Fold llvm.amdgcn.cube* intrinsics
ClosedPublic

Authored by foad on Feb 26 2020, 8:27 AM.

Details

Reviewers
nhaehnle
arsenm
rampitec
Commits
rG596446623bad: [AMDGPU][ConstantFolding] Fold llvm.amdgcn.cube* intrinsics
Summary

This folds the following family of intrinsics:
llvm.amdgcn.cubeid (face id)
llvm.amdgcn.cubema (major axis)
llvm.amdgcn.cubesc (S coordinate)
llvm.amdgcn.cubetc (T coordinate)

Diff Detail

Event Timeline

foad created this revision.Feb 26 2020, 8:27 AM
Herald added a project: Restricted Project. · View Herald TranscriptFeb 26 2020, 8:27 AM
Herald added subscribers: llvm-commits, kerbowa, hiraditya and 7 others. · View Herald Transcript
arsenm added inline comments.Feb 26 2020, 8:41 AM
llvm/lib/Analysis/ConstantFolding.cpp
2312

Using host FP operations is generally frowned on, and this should use APFloat

foad marked an inline comment as done.Feb 26 2020, 8:49 AM
foad added inline comments.
llvm/lib/Analysis/ConstantFolding.cpp
2312

OK, but the APFloat API is horrible to use for anything more complicated than + - * /.

Harbormaster failed remote builds in B47317: Diff 246744!Feb 26 2020, 8:51 AM
foad updated this revision to Diff 246824.Feb 26 2020, 1:19 PM

Use APFloats.

Harbormaster failed remote builds in B47359: Diff 246824!Feb 26 2020, 1:42 PM
arsenm added inline comments.Feb 26 2020, 1:53 PM
llvm/lib/Analysis/ConstantFolding.cpp
2313

const references

2319–2322

This can use const references and can be moved to a separate function since there's no local captures

2378–2380

Avoid the copy by sinking thees to the uses

foad updated this revision to Diff 246909.Feb 27 2020, 4:16 AM

Address review comments, rebase on D75237.

foad added a parent revision: D75237: [APFloat] Overload comparison operators.Feb 27 2020, 4:16 AM
foad marked 5 inline comments as done.
foad added inline comments.
llvm/lib/Analysis/ConstantFolding.cpp
2319–2322

I moved it to APFloat::operator>= instead.

Harbormaster failed remote builds in B47401: Diff 246909!Feb 27 2020, 4:52 AM
arsenm accepted this revision.Mar 4 2020, 3:54 PM

LGTM

This revision is now accepted and ready to land.Mar 4 2020, 3:54 PM
Closed by commit rG596446623bad: [AMDGPU][ConstantFolding] Fold llvm.amdgcn.cube* intrinsics (authored by foad). · Explain WhyMar 6 2020, 8:49 AM
This revision was automatically updated to reflect the committed changes.
foad marked an inline comment as done.

Revision Contents

PathSize
llvm/
lib/
Analysis/
68 lines
test/
Analysis/
ConstantFolding/
AMDGPU/
155 lines
155 lines
155 lines
155 lines

Diff 248753

llvm/lib/Analysis/ConstantFolding.cpp

Show First 20 Lines • Show All 1,441 Lines • ▼ Show 20 Linesbool llvm::canConstantFoldCallTo(const CallBase *Call, const Function *F) {
case Intrinsic::uadd_sat: case Intrinsic::uadd_sat:
case Intrinsic::ssub_sat: case Intrinsic::ssub_sat:
case Intrinsic::usub_sat: case Intrinsic::usub_sat:
case Intrinsic::smul_fix: case Intrinsic::smul_fix:
case Intrinsic::smul_fix_sat: case Intrinsic::smul_fix_sat:
case Intrinsic::convert_from_fp16: case Intrinsic::convert_from_fp16:
case Intrinsic::convert_to_fp16: case Intrinsic::convert_to_fp16:
case Intrinsic::bitreverse: case Intrinsic::bitreverse:
case Intrinsic::amdgcn_cubeid:
case Intrinsic::amdgcn_cubema:
case Intrinsic::amdgcn_cubesc:
case Intrinsic::amdgcn_cubetc:
case Intrinsic::amdgcn_fmul_legacy: case Intrinsic::amdgcn_fmul_legacy:
case Intrinsic::amdgcn_fract: case Intrinsic::amdgcn_fract:
case Intrinsic::x86_sse_cvtss2si: case Intrinsic::x86_sse_cvtss2si:
case Intrinsic::x86_sse_cvtss2si64: case Intrinsic::x86_sse_cvtss2si64:
case Intrinsic::x86_sse_cvttss2si: case Intrinsic::x86_sse_cvttss2si:
case Intrinsic::x86_sse_cvttss2si64: case Intrinsic::x86_sse_cvttss2si64:
case Intrinsic::x86_sse2_cvtsd2si: case Intrinsic::x86_sse2_cvtsd2si:
case Intrinsic::x86_sse2_cvtsd2si64: case Intrinsic::x86_sse2_cvtsd2si64:
▲ Show 20 Lines • Show All 842 Lines • ▼ Show 20 Lines case Intrinsic::x86_avx512_cvttsd2usi64:
/*roundTowardZero=*/true, Ty, /*roundTowardZero=*/true, Ty,
/*IsSigned*/false); /*IsSigned*/false);
break; break;
} }
} }
return nullptr; return nullptr;
} }
static APFloat ConstantFoldAMDGCNCubeIntrinsic(Intrinsic::ID IntrinsicID,
arsenmUnsubmitted
Done
ReplyInline Actions

Using host FP operations is generally frowned on, and this should use APFloat

arsenm: Using host FP operations is generally frowned on, and this should use APFloat
foadAuthorUnsubmitted
Done
ReplyInline Actions

OK, but the APFloat API is horrible to use for anything more complicated than + - * /.

foad: OK, but the APFloat API is horrible to use for anything more complicated than + - * /.
const APFloat &S0,
arsenmUnsubmitted
Done
ReplyInline Actions

const references

arsenm: const references
const APFloat &S1,
const APFloat &S2) {
unsigned ID;
const fltSemantics &Sem = S0.getSemantics();
APFloat MA(Sem), SC(Sem), TC(Sem);
if (abs(S2) >= abs(S0) && abs(S2) >= abs(S1)) {
if (S2.isNegative() && S2.isNonZero() && !S2.isNaN()) {
// S2 < 0
ID = 5;
arsenmUnsubmitted
Done
ReplyInline Actions

This can use const references and can be moved to a separate function since there's no local captures

arsenm: This can use const references and can be moved to a separate function since there's no local…
foadAuthorUnsubmitted
Done
ReplyInline Actions

I moved it to APFloat::operator>= instead.

foad: I moved it to APFloat::operator>= instead.
SC = -S0;
} else {
ID = 4;
SC = S0;
}
MA = S2;
TC = -S1;
} else if (abs(S1) >= abs(S0)) {
if (S1.isNegative() && S1.isNonZero() && !S1.isNaN()) {
// S1 < 0
ID = 3;
TC = -S2;
} else {
ID = 2;
TC = S2;
}
MA = S1;
SC = S0;
} else {
if (S0.isNegative() && S0.isNonZero() && !S0.isNaN()) {
// S0 < 0
ID = 1;
SC = S2;
} else {
ID = 0;
SC = -S2;
}
MA = S0;
TC = -S1;
}
switch (IntrinsicID) {
default:
llvm_unreachable("unhandled amdgcn cube intrinsic");
case Intrinsic::amdgcn_cubeid:
return APFloat(Sem, ID);
case Intrinsic::amdgcn_cubema:
return MA + MA;
case Intrinsic::amdgcn_cubesc:
return SC;
case Intrinsic::amdgcn_cubetc:
return TC;
}
}
static Constant *ConstantFoldScalarCall3(StringRef Name, static Constant *ConstantFoldScalarCall3(StringRef Name,
Intrinsic::ID IntrinsicID, Intrinsic::ID IntrinsicID,
Type *Ty, Type *Ty,
ArrayRef<Constant *> Operands, ArrayRef<Constant *> Operands,
const TargetLibraryInfo *TLI, const TargetLibraryInfo *TLI,
const CallBase *Call) { const CallBase *Call) {
assert(Operands.size() == 3 && "Wrong number of operands."); assert(Operands.size() == 3 && "Wrong number of operands.");
if (const auto *Op1 = dyn_cast<ConstantFP>(Operands[0])) { if (const auto *Op1 = dyn_cast<ConstantFP>(Operands[0])) {
if (const auto *Op2 = dyn_cast<ConstantFP>(Operands[1])) { if (const auto *Op2 = dyn_cast<ConstantFP>(Operands[1])) {
if (const auto *Op3 = dyn_cast<ConstantFP>(Operands[2])) { if (const auto *Op3 = dyn_cast<ConstantFP>(Operands[2])) {
switch (IntrinsicID) { switch (IntrinsicID) {
default: break; default: break;
case Intrinsic::fma: case Intrinsic::fma:
arsenmUnsubmitted
Done
ReplyInline Actions

Avoid the copy by sinking thees to the uses

arsenm: Avoid the copy by sinking thees to the uses
case Intrinsic::fmuladd: { case Intrinsic::fmuladd: {
APFloat V = Op1->getValueAPF(); APFloat V = Op1->getValueAPF();
V.fusedMultiplyAdd(Op2->getValueAPF(), Op3->getValueAPF(), V.fusedMultiplyAdd(Op2->getValueAPF(), Op3->getValueAPF(),
APFloat::rmNearestTiesToEven); APFloat::rmNearestTiesToEven);
return ConstantFP::get(Ty->getContext(), V); return ConstantFP::get(Ty->getContext(), V);
} }
case Intrinsic::amdgcn_cubeid:
case Intrinsic::amdgcn_cubema:
case Intrinsic::amdgcn_cubesc:
case Intrinsic::amdgcn_cubetc: {
APFloat V = ConstantFoldAMDGCNCubeIntrinsic(
IntrinsicID, Op1->getValueAPF(), Op2->getValueAPF(),
Op3->getValueAPF());
return ConstantFP::get(Ty->getContext(), V);
}
} }
} }
} }
} }
if (const auto *Op1 = dyn_cast<ConstantInt>(Operands[0])) { if (const auto *Op1 = dyn_cast<ConstantInt>(Operands[0])) {
if (const auto *Op2 = dyn_cast<ConstantInt>(Operands[1])) { if (const auto *Op2 = dyn_cast<ConstantInt>(Operands[1])) {
if (const auto *Op3 = dyn_cast<ConstantInt>(Operands[2])) { if (const auto *Op3 = dyn_cast<ConstantInt>(Operands[2])) {
▲ Show 20 Lines • Show All 336 Lines • Show Last 20 Lines

llvm/test/Analysis/ConstantFolding/AMDGPU/cubeid.ll

  • This file was added.
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -instsimplify -S | FileCheck %s
declare float @llvm.amdgcn.cubeid(float, float, float)
define void @test(float* %p) {
; CHECK-LABEL: @test(
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P:%.*]]
; CHECK-NEXT: store volatile float 2.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 2.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 0.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 0.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 5.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 2.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 5.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 2.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 0.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 0.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 0.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 0.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 5.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 5.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 0.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 0.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 2.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 2.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 5.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 2.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 5.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 2.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 5.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 5.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+00, float* [[P]]
; CHECK-NEXT: ret void
;
%p3p4p5 = call float @llvm.amdgcn.cubeid(float +3.0, float +4.0, float +5.0)
store volatile float %p3p4p5, float* %p
%p3p5p4 = call float @llvm.amdgcn.cubeid(float +3.0, float +5.0, float +4.0)
store volatile float %p3p5p4, float* %p
%p4p3p5 = call float @llvm.amdgcn.cubeid(float +4.0, float +3.0, float +5.0)
store volatile float %p4p3p5, float* %p
%p4p5p3 = call float @llvm.amdgcn.cubeid(float +4.0, float +5.0, float +3.0)
store volatile float %p4p5p3, float* %p
%p5p3p4 = call float @llvm.amdgcn.cubeid(float +5.0, float +3.0, float +4.0)
store volatile float %p5p3p4, float* %p
%p5p4p3 = call float @llvm.amdgcn.cubeid(float +5.0, float +4.0, float +3.0)
store volatile float %p5p4p3, float* %p
%p3p4n5 = call float @llvm.amdgcn.cubeid(float +3.0, float +4.0, float -5.0)
store volatile float %p3p4n5, float* %p
%p3p5n4 = call float @llvm.amdgcn.cubeid(float +3.0, float +5.0, float -4.0)
store volatile float %p3p5n4, float* %p
%p4p3n5 = call float @llvm.amdgcn.cubeid(float +4.0, float +3.0, float -5.0)
store volatile float %p4p3n5, float* %p
%p4p5n3 = call float @llvm.amdgcn.cubeid(float +4.0, float +5.0, float -3.0)
store volatile float %p4p5n3, float* %p
%p5p3n4 = call float @llvm.amdgcn.cubeid(float +5.0, float +3.0, float -4.0)
store volatile float %p5p3n4, float* %p
%p5p4n3 = call float @llvm.amdgcn.cubeid(float +5.0, float +4.0, float -3.0)
store volatile float %p5p4n3, float* %p
%p3n4p5 = call float @llvm.amdgcn.cubeid(float +3.0, float -4.0, float +5.0)
store volatile float %p3n4p5, float* %p
%p3n5p4 = call float @llvm.amdgcn.cubeid(float +3.0, float -5.0, float +4.0)
store volatile float %p3n5p4, float* %p
%p4n3p5 = call float @llvm.amdgcn.cubeid(float +4.0, float -3.0, float +5.0)
store volatile float %p4n3p5, float* %p
%p4n5p3 = call float @llvm.amdgcn.cubeid(float +4.0, float -5.0, float +3.0)
store volatile float %p4n5p3, float* %p
%p5n3p4 = call float @llvm.amdgcn.cubeid(float +5.0, float -3.0, float +4.0)
store volatile float %p5n3p4, float* %p
%p5n4p3 = call float @llvm.amdgcn.cubeid(float +5.0, float -4.0, float +3.0)
store volatile float %p5n4p3, float* %p
%p3n4n5 = call float @llvm.amdgcn.cubeid(float +3.0, float -4.0, float -5.0)
store volatile float %p3n4n5, float* %p
%p3n5n4 = call float @llvm.amdgcn.cubeid(float +3.0, float -5.0, float -4.0)
store volatile float %p3n5n4, float* %p
%p4n3n5 = call float @llvm.amdgcn.cubeid(float +4.0, float -3.0, float -5.0)
store volatile float %p4n3n5, float* %p
%p4n5n3 = call float @llvm.amdgcn.cubeid(float +4.0, float -5.0, float -3.0)
store volatile float %p4n5n3, float* %p
%p5n3n4 = call float @llvm.amdgcn.cubeid(float +5.0, float -3.0, float -4.0)
store volatile float %p5n3n4, float* %p
%p5n4n3 = call float @llvm.amdgcn.cubeid(float +5.0, float -4.0, float -3.0)
store volatile float %p5n4n3, float* %p
%n3p4p5 = call float @llvm.amdgcn.cubeid(float -3.0, float +4.0, float +5.0)
store volatile float %n3p4p5, float* %p
%n3p5p4 = call float @llvm.amdgcn.cubeid(float -3.0, float +5.0, float +4.0)
store volatile float %n3p5p4, float* %p
%n4p3p5 = call float @llvm.amdgcn.cubeid(float -4.0, float +3.0, float +5.0)
store volatile float %n4p3p5, float* %p
%n4p5p3 = call float @llvm.amdgcn.cubeid(float -4.0, float +5.0, float +3.0)
store volatile float %n4p5p3, float* %p
%n5p3p4 = call float @llvm.amdgcn.cubeid(float -5.0, float +3.0, float +4.0)
store volatile float %n5p3p4, float* %p
%n5p4p3 = call float @llvm.amdgcn.cubeid(float -5.0, float +4.0, float +3.0)
store volatile float %n5p4p3, float* %p
%n3p4n5 = call float @llvm.amdgcn.cubeid(float -3.0, float +4.0, float -5.0)
store volatile float %n3p4n5, float* %p
%n3p5n4 = call float @llvm.amdgcn.cubeid(float -3.0, float +5.0, float -4.0)
store volatile float %n3p5n4, float* %p
%n4p3n5 = call float @llvm.amdgcn.cubeid(float -4.0, float +3.0, float -5.0)
store volatile float %n4p3n5, float* %p
%n4p5n3 = call float @llvm.amdgcn.cubeid(float -4.0, float +5.0, float -3.0)
store volatile float %n4p5n3, float* %p
%n5p3n4 = call float @llvm.amdgcn.cubeid(float -5.0, float +3.0, float -4.0)
store volatile float %n5p3n4, float* %p
%n5p4n3 = call float @llvm.amdgcn.cubeid(float -5.0, float +4.0, float -3.0)
store volatile float %n5p4n3, float* %p
%n3n4p5 = call float @llvm.amdgcn.cubeid(float -3.0, float -4.0, float +5.0)
store volatile float %n3n4p5, float* %p
%n3n5p4 = call float @llvm.amdgcn.cubeid(float -3.0, float -5.0, float +4.0)
store volatile float %n3n5p4, float* %p
%n4n3p5 = call float @llvm.amdgcn.cubeid(float -4.0, float -3.0, float +5.0)
store volatile float %n4n3p5, float* %p
%n4n5p3 = call float @llvm.amdgcn.cubeid(float -4.0, float -5.0, float +3.0)
store volatile float %n4n5p3, float* %p
%n5n3p4 = call float @llvm.amdgcn.cubeid(float -5.0, float -3.0, float +4.0)
store volatile float %n5n3p4, float* %p
%n5n4p3 = call float @llvm.amdgcn.cubeid(float -5.0, float -4.0, float +3.0)
store volatile float %n5n4p3, float* %p
%n3n4n5 = call float @llvm.amdgcn.cubeid(float -3.0, float -4.0, float -5.0)
store volatile float %n3n4n5, float* %p
%n3n5n4 = call float @llvm.amdgcn.cubeid(float -3.0, float -5.0, float -4.0)
store volatile float %n3n5n4, float* %p
%n4n3n5 = call float @llvm.amdgcn.cubeid(float -4.0, float -3.0, float -5.0)
store volatile float %n4n3n5, float* %p
%n4n5n3 = call float @llvm.amdgcn.cubeid(float -4.0, float -5.0, float -3.0)
store volatile float %n4n5n3, float* %p
%n5n3n4 = call float @llvm.amdgcn.cubeid(float -5.0, float -3.0, float -4.0)
store volatile float %n5n3n4, float* %p
%n5n4n3 = call float @llvm.amdgcn.cubeid(float -5.0, float -4.0, float -3.0)
store volatile float %n5n4n3, float* %p
ret void
}

llvm/test/Analysis/ConstantFolding/AMDGPU/cubema.ll

  • This file was added.
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -instsimplify -S | FileCheck %s
declare float @llvm.amdgcn.cubema(float, float, float)
define void @test(float* %p) {
; CHECK-LABEL: @test(
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P:%.*]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float 1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: store volatile float -1.000000e+01, float* [[P]]
; CHECK-NEXT: ret void
;
%p3p4p5 = call float @llvm.amdgcn.cubema(float +3.0, float +4.0, float +5.0)
store volatile float %p3p4p5, float* %p
%p3p5p4 = call float @llvm.amdgcn.cubema(float +3.0, float +5.0, float +4.0)
store volatile float %p3p5p4, float* %p
%p4p3p5 = call float @llvm.amdgcn.cubema(float +4.0, float +3.0, float +5.0)
store volatile float %p4p3p5, float* %p
%p4p5p3 = call float @llvm.amdgcn.cubema(float +4.0, float +5.0, float +3.0)
store volatile float %p4p5p3, float* %p
%p5p3p4 = call float @llvm.amdgcn.cubema(float +5.0, float +3.0, float +4.0)
store volatile float %p5p3p4, float* %p
%p5p4p3 = call float @llvm.amdgcn.cubema(float +5.0, float +4.0, float +3.0)
store volatile float %p5p4p3, float* %p
%p3p4n5 = call float @llvm.amdgcn.cubema(float +3.0, float +4.0, float -5.0)
store volatile float %p3p4n5, float* %p
%p3p5n4 = call float @llvm.amdgcn.cubema(float +3.0, float +5.0, float -4.0)
store volatile float %p3p5n4, float* %p
%p4p3n5 = call float @llvm.amdgcn.cubema(float +4.0, float +3.0, float -5.0)
store volatile float %p4p3n5, float* %p
%p4p5n3 = call float @llvm.amdgcn.cubema(float +4.0, float +5.0, float -3.0)
store volatile float %p4p5n3, float* %p
%p5p3n4 = call float @llvm.amdgcn.cubema(float +5.0, float +3.0, float -4.0)
store volatile float %p5p3n4, float* %p
%p5p4n3 = call float @llvm.amdgcn.cubema(float +5.0, float +4.0, float -3.0)
store volatile float %p5p4n3, float* %p
%p3n4p5 = call float @llvm.amdgcn.cubema(float +3.0, float -4.0, float +5.0)
store volatile float %p3n4p5, float* %p
%p3n5p4 = call float @llvm.amdgcn.cubema(float +3.0, float -5.0, float +4.0)
store volatile float %p3n5p4, float* %p
%p4n3p5 = call float @llvm.amdgcn.cubema(float +4.0, float -3.0, float +5.0)
store volatile float %p4n3p5, float* %p
%p4n5p3 = call float @llvm.amdgcn.cubema(float +4.0, float -5.0, float +3.0)
store volatile float %p4n5p3, float* %p
%p5n3p4 = call float @llvm.amdgcn.cubema(float +5.0, float -3.0, float +4.0)
store volatile float %p5n3p4, float* %p
%p5n4p3 = call float @llvm.amdgcn.cubema(float +5.0, float -4.0, float +3.0)
store volatile float %p5n4p3, float* %p
%p3n4n5 = call float @llvm.amdgcn.cubema(float +3.0, float -4.0, float -5.0)
store volatile float %p3n4n5, float* %p
%p3n5n4 = call float @llvm.amdgcn.cubema(float +3.0, float -5.0, float -4.0)
store volatile float %p3n5n4, float* %p
%p4n3n5 = call float @llvm.amdgcn.cubema(float +4.0, float -3.0, float -5.0)
store volatile float %p4n3n5, float* %p
%p4n5n3 = call float @llvm.amdgcn.cubema(float +4.0, float -5.0, float -3.0)
store volatile float %p4n5n3, float* %p
%p5n3n4 = call float @llvm.amdgcn.cubema(float +5.0, float -3.0, float -4.0)
store volatile float %p5n3n4, float* %p
%p5n4n3 = call float @llvm.amdgcn.cubema(float +5.0, float -4.0, float -3.0)
store volatile float %p5n4n3, float* %p
%n3p4p5 = call float @llvm.amdgcn.cubema(float -3.0, float +4.0, float +5.0)
store volatile float %n3p4p5, float* %p
%n3p5p4 = call float @llvm.amdgcn.cubema(float -3.0, float +5.0, float +4.0)
store volatile float %n3p5p4, float* %p
%n4p3p5 = call float @llvm.amdgcn.cubema(float -4.0, float +3.0, float +5.0)
store volatile float %n4p3p5, float* %p
%n4p5p3 = call float @llvm.amdgcn.cubema(float -4.0, float +5.0, float +3.0)
store volatile float %n4p5p3, float* %p
%n5p3p4 = call float @llvm.amdgcn.cubema(float -5.0, float +3.0, float +4.0)
store volatile float %n5p3p4, float* %p
%n5p4p3 = call float @llvm.amdgcn.cubema(float -5.0, float +4.0, float +3.0)
store volatile float %n5p4p3, float* %p
%n3p4n5 = call float @llvm.amdgcn.cubema(float -3.0, float +4.0, float -5.0)
store volatile float %n3p4n5, float* %p
%n3p5n4 = call float @llvm.amdgcn.cubema(float -3.0, float +5.0, float -4.0)
store volatile float %n3p5n4, float* %p
%n4p3n5 = call float @llvm.amdgcn.cubema(float -4.0, float +3.0, float -5.0)
store volatile float %n4p3n5, float* %p
%n4p5n3 = call float @llvm.amdgcn.cubema(float -4.0, float +5.0, float -3.0)
store volatile float %n4p5n3, float* %p
%n5p3n4 = call float @llvm.amdgcn.cubema(float -5.0, float +3.0, float -4.0)
store volatile float %n5p3n4, float* %p
%n5p4n3 = call float @llvm.amdgcn.cubema(float -5.0, float +4.0, float -3.0)
store volatile float %n5p4n3, float* %p
%n3n4p5 = call float @llvm.amdgcn.cubema(float -3.0, float -4.0, float +5.0)
store volatile float %n3n4p5, float* %p
%n3n5p4 = call float @llvm.amdgcn.cubema(float -3.0, float -5.0, float +4.0)
store volatile float %n3n5p4, float* %p
%n4n3p5 = call float @llvm.amdgcn.cubema(float -4.0, float -3.0, float +5.0)
store volatile float %n4n3p5, float* %p
%n4n5p3 = call float @llvm.amdgcn.cubema(float -4.0, float -5.0, float +3.0)
store volatile float %n4n5p3, float* %p
%n5n3p4 = call float @llvm.amdgcn.cubema(float -5.0, float -3.0, float +4.0)
store volatile float %n5n3p4, float* %p
%n5n4p3 = call float @llvm.amdgcn.cubema(float -5.0, float -4.0, float +3.0)
store volatile float %n5n4p3, float* %p
%n3n4n5 = call float @llvm.amdgcn.cubema(float -3.0, float -4.0, float -5.0)
store volatile float %n3n4n5, float* %p
%n3n5n4 = call float @llvm.amdgcn.cubema(float -3.0, float -5.0, float -4.0)
store volatile float %n3n5n4, float* %p
%n4n3n5 = call float @llvm.amdgcn.cubema(float -4.0, float -3.0, float -5.0)
store volatile float %n4n3n5, float* %p
%n4n5n3 = call float @llvm.amdgcn.cubema(float -4.0, float -5.0, float -3.0)
store volatile float %n4n5n3, float* %p
%n5n3n4 = call float @llvm.amdgcn.cubema(float -5.0, float -3.0, float -4.0)
store volatile float %n5n3n4, float* %p
%n5n4n3 = call float @llvm.amdgcn.cubema(float -5.0, float -4.0, float -3.0)
store volatile float %n5n4n3, float* %p
ret void
}

llvm/test/Analysis/ConstantFolding/AMDGPU/cubesc.ll

  • This file was added.
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -instsimplify -S | FileCheck %s
declare float @llvm.amdgcn.cubesc(float, float, float)
define void @test(float* %p) {
; CHECK-LABEL: @test(
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P:%.*]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: ret void
;
%p3p4p5 = call float @llvm.amdgcn.cubesc(float +3.0, float +4.0, float +5.0)
store volatile float %p3p4p5, float* %p
%p3p5p4 = call float @llvm.amdgcn.cubesc(float +3.0, float +5.0, float +4.0)
store volatile float %p3p5p4, float* %p
%p4p3p5 = call float @llvm.amdgcn.cubesc(float +4.0, float +3.0, float +5.0)
store volatile float %p4p3p5, float* %p
%p4p5p3 = call float @llvm.amdgcn.cubesc(float +4.0, float +5.0, float +3.0)
store volatile float %p4p5p3, float* %p
%p5p3p4 = call float @llvm.amdgcn.cubesc(float +5.0, float +3.0, float +4.0)
store volatile float %p5p3p4, float* %p
%p5p4p3 = call float @llvm.amdgcn.cubesc(float +5.0, float +4.0, float +3.0)
store volatile float %p5p4p3, float* %p
%p3p4n5 = call float @llvm.amdgcn.cubesc(float +3.0, float +4.0, float -5.0)
store volatile float %p3p4n5, float* %p
%p3p5n4 = call float @llvm.amdgcn.cubesc(float +3.0, float +5.0, float -4.0)
store volatile float %p3p5n4, float* %p
%p4p3n5 = call float @llvm.amdgcn.cubesc(float +4.0, float +3.0, float -5.0)
store volatile float %p4p3n5, float* %p
%p4p5n3 = call float @llvm.amdgcn.cubesc(float +4.0, float +5.0, float -3.0)
store volatile float %p4p5n3, float* %p
%p5p3n4 = call float @llvm.amdgcn.cubesc(float +5.0, float +3.0, float -4.0)
store volatile float %p5p3n4, float* %p
%p5p4n3 = call float @llvm.amdgcn.cubesc(float +5.0, float +4.0, float -3.0)
store volatile float %p5p4n3, float* %p
%p3n4p5 = call float @llvm.amdgcn.cubesc(float +3.0, float -4.0, float +5.0)
store volatile float %p3n4p5, float* %p
%p3n5p4 = call float @llvm.amdgcn.cubesc(float +3.0, float -5.0, float +4.0)
store volatile float %p3n5p4, float* %p
%p4n3p5 = call float @llvm.amdgcn.cubesc(float +4.0, float -3.0, float +5.0)
store volatile float %p4n3p5, float* %p
%p4n5p3 = call float @llvm.amdgcn.cubesc(float +4.0, float -5.0, float +3.0)
store volatile float %p4n5p3, float* %p
%p5n3p4 = call float @llvm.amdgcn.cubesc(float +5.0, float -3.0, float +4.0)
store volatile float %p5n3p4, float* %p
%p5n4p3 = call float @llvm.amdgcn.cubesc(float +5.0, float -4.0, float +3.0)
store volatile float %p5n4p3, float* %p
%p3n4n5 = call float @llvm.amdgcn.cubesc(float +3.0, float -4.0, float -5.0)
store volatile float %p3n4n5, float* %p
%p3n5n4 = call float @llvm.amdgcn.cubesc(float +3.0, float -5.0, float -4.0)
store volatile float %p3n5n4, float* %p
%p4n3n5 = call float @llvm.amdgcn.cubesc(float +4.0, float -3.0, float -5.0)
store volatile float %p4n3n5, float* %p
%p4n5n3 = call float @llvm.amdgcn.cubesc(float +4.0, float -5.0, float -3.0)
store volatile float %p4n5n3, float* %p
%p5n3n4 = call float @llvm.amdgcn.cubesc(float +5.0, float -3.0, float -4.0)
store volatile float %p5n3n4, float* %p
%p5n4n3 = call float @llvm.amdgcn.cubesc(float +5.0, float -4.0, float -3.0)
store volatile float %p5n4n3, float* %p
%n3p4p5 = call float @llvm.amdgcn.cubesc(float -3.0, float +4.0, float +5.0)
store volatile float %n3p4p5, float* %p
%n3p5p4 = call float @llvm.amdgcn.cubesc(float -3.0, float +5.0, float +4.0)
store volatile float %n3p5p4, float* %p
%n4p3p5 = call float @llvm.amdgcn.cubesc(float -4.0, float +3.0, float +5.0)
store volatile float %n4p3p5, float* %p
%n4p5p3 = call float @llvm.amdgcn.cubesc(float -4.0, float +5.0, float +3.0)
store volatile float %n4p5p3, float* %p
%n5p3p4 = call float @llvm.amdgcn.cubesc(float -5.0, float +3.0, float +4.0)
store volatile float %n5p3p4, float* %p
%n5p4p3 = call float @llvm.amdgcn.cubesc(float -5.0, float +4.0, float +3.0)
store volatile float %n5p4p3, float* %p
%n3p4n5 = call float @llvm.amdgcn.cubesc(float -3.0, float +4.0, float -5.0)
store volatile float %n3p4n5, float* %p
%n3p5n4 = call float @llvm.amdgcn.cubesc(float -3.0, float +5.0, float -4.0)
store volatile float %n3p5n4, float* %p
%n4p3n5 = call float @llvm.amdgcn.cubesc(float -4.0, float +3.0, float -5.0)
store volatile float %n4p3n5, float* %p
%n4p5n3 = call float @llvm.amdgcn.cubesc(float -4.0, float +5.0, float -3.0)
store volatile float %n4p5n3, float* %p
%n5p3n4 = call float @llvm.amdgcn.cubesc(float -5.0, float +3.0, float -4.0)
store volatile float %n5p3n4, float* %p
%n5p4n3 = call float @llvm.amdgcn.cubesc(float -5.0, float +4.0, float -3.0)
store volatile float %n5p4n3, float* %p
%n3n4p5 = call float @llvm.amdgcn.cubesc(float -3.0, float -4.0, float +5.0)
store volatile float %n3n4p5, float* %p
%n3n5p4 = call float @llvm.amdgcn.cubesc(float -3.0, float -5.0, float +4.0)
store volatile float %n3n5p4, float* %p
%n4n3p5 = call float @llvm.amdgcn.cubesc(float -4.0, float -3.0, float +5.0)
store volatile float %n4n3p5, float* %p
%n4n5p3 = call float @llvm.amdgcn.cubesc(float -4.0, float -5.0, float +3.0)
store volatile float %n4n5p3, float* %p
%n5n3p4 = call float @llvm.amdgcn.cubesc(float -5.0, float -3.0, float +4.0)
store volatile float %n5n3p4, float* %p
%n5n4p3 = call float @llvm.amdgcn.cubesc(float -5.0, float -4.0, float +3.0)
store volatile float %n5n4p3, float* %p
%n3n4n5 = call float @llvm.amdgcn.cubesc(float -3.0, float -4.0, float -5.0)
store volatile float %n3n4n5, float* %p
%n3n5n4 = call float @llvm.amdgcn.cubesc(float -3.0, float -5.0, float -4.0)
store volatile float %n3n5n4, float* %p
%n4n3n5 = call float @llvm.amdgcn.cubesc(float -4.0, float -3.0, float -5.0)
store volatile float %n4n3n5, float* %p
%n4n5n3 = call float @llvm.amdgcn.cubesc(float -4.0, float -5.0, float -3.0)
store volatile float %n4n5n3, float* %p
%n5n3n4 = call float @llvm.amdgcn.cubesc(float -5.0, float -3.0, float -4.0)
store volatile float %n5n3n4, float* %p
%n5n4n3 = call float @llvm.amdgcn.cubesc(float -5.0, float -4.0, float -3.0)
store volatile float %n5n4n3, float* %p
ret void
}

llvm/test/Analysis/ConstantFolding/AMDGPU/cubetc.ll

  • This file was added.
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -instsimplify -S | FileCheck %s
declare float @llvm.amdgcn.cubetc(float, float, float)
define void @test(float* %p) {
; CHECK-LABEL: @test(
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P:%.*]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float -3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 3.000000e+00, float* [[P]]
; CHECK-NEXT: store volatile float 4.000000e+00, float* [[P]]
; CHECK-NEXT: ret void
;
%p3p4p5 = call float @llvm.amdgcn.cubetc(float +3.0, float +4.0, float +5.0)
store volatile float %p3p4p5, float* %p
%p3p5p4 = call float @llvm.amdgcn.cubetc(float +3.0, float +5.0, float +4.0)
store volatile float %p3p5p4, float* %p
%p4p3p5 = call float @llvm.amdgcn.cubetc(float +4.0, float +3.0, float +5.0)
store volatile float %p4p3p5, float* %p
%p4p5p3 = call float @llvm.amdgcn.cubetc(float +4.0, float +5.0, float +3.0)
store volatile float %p4p5p3, float* %p
%p5p3p4 = call float @llvm.amdgcn.cubetc(float +5.0, float +3.0, float +4.0)
store volatile float %p5p3p4, float* %p
%p5p4p3 = call float @llvm.amdgcn.cubetc(float +5.0, float +4.0, float +3.0)
store volatile float %p5p4p3, float* %p
%p3p4n5 = call float @llvm.amdgcn.cubetc(float +3.0, float +4.0, float -5.0)
store volatile float %p3p4n5, float* %p
%p3p5n4 = call float @llvm.amdgcn.cubetc(float +3.0, float +5.0, float -4.0)
store volatile float %p3p5n4, float* %p
%p4p3n5 = call float @llvm.amdgcn.cubetc(float +4.0, float +3.0, float -5.0)
store volatile float %p4p3n5, float* %p
%p4p5n3 = call float @llvm.amdgcn.cubetc(float +4.0, float +5.0, float -3.0)
store volatile float %p4p5n3, float* %p
%p5p3n4 = call float @llvm.amdgcn.cubetc(float +5.0, float +3.0, float -4.0)
store volatile float %p5p3n4, float* %p
%p5p4n3 = call float @llvm.amdgcn.cubetc(float +5.0, float +4.0, float -3.0)
store volatile float %p5p4n3, float* %p
%p3n4p5 = call float @llvm.amdgcn.cubetc(float +3.0, float -4.0, float +5.0)
store volatile float %p3n4p5, float* %p
%p3n5p4 = call float @llvm.amdgcn.cubetc(float +3.0, float -5.0, float +4.0)
store volatile float %p3n5p4, float* %p
%p4n3p5 = call float @llvm.amdgcn.cubetc(float +4.0, float -3.0, float +5.0)
store volatile float %p4n3p5, float* %p
%p4n5p3 = call float @llvm.amdgcn.cubetc(float +4.0, float -5.0, float +3.0)
store volatile float %p4n5p3, float* %p
%p5n3p4 = call float @llvm.amdgcn.cubetc(float +5.0, float -3.0, float +4.0)
store volatile float %p5n3p4, float* %p
%p5n4p3 = call float @llvm.amdgcn.cubetc(float +5.0, float -4.0, float +3.0)
store volatile float %p5n4p3, float* %p
%p3n4n5 = call float @llvm.amdgcn.cubetc(float +3.0, float -4.0, float -5.0)
store volatile float %p3n4n5, float* %p
%p3n5n4 = call float @llvm.amdgcn.cubetc(float +3.0, float -5.0, float -4.0)
store volatile float %p3n5n4, float* %p
%p4n3n5 = call float @llvm.amdgcn.cubetc(float +4.0, float -3.0, float -5.0)
store volatile float %p4n3n5, float* %p
%p4n5n3 = call float @llvm.amdgcn.cubetc(float +4.0, float -5.0, float -3.0)
store volatile float %p4n5n3, float* %p
%p5n3n4 = call float @llvm.amdgcn.cubetc(float +5.0, float -3.0, float -4.0)
store volatile float %p5n3n4, float* %p
%p5n4n3 = call float @llvm.amdgcn.cubetc(float +5.0, float -4.0, float -3.0)
store volatile float %p5n4n3, float* %p
%n3p4p5 = call float @llvm.amdgcn.cubetc(float -3.0, float +4.0, float +5.0)
store volatile float %n3p4p5, float* %p
%n3p5p4 = call float @llvm.amdgcn.cubetc(float -3.0, float +5.0, float +4.0)
store volatile float %n3p5p4, float* %p
%n4p3p5 = call float @llvm.amdgcn.cubetc(float -4.0, float +3.0, float +5.0)
store volatile float %n4p3p5, float* %p
%n4p5p3 = call float @llvm.amdgcn.cubetc(float -4.0, float +5.0, float +3.0)
store volatile float %n4p5p3, float* %p
%n5p3p4 = call float @llvm.amdgcn.cubetc(float -5.0, float +3.0, float +4.0)
store volatile float %n5p3p4, float* %p
%n5p4p3 = call float @llvm.amdgcn.cubetc(float -5.0, float +4.0, float +3.0)
store volatile float %n5p4p3, float* %p
%n3p4n5 = call float @llvm.amdgcn.cubetc(float -3.0, float +4.0, float -5.0)
store volatile float %n3p4n5, float* %p
%n3p5n4 = call float @llvm.amdgcn.cubetc(float -3.0, float +5.0, float -4.0)
store volatile float %n3p5n4, float* %p
%n4p3n5 = call float @llvm.amdgcn.cubetc(float -4.0, float +3.0, float -5.0)
store volatile float %n4p3n5, float* %p
%n4p5n3 = call float @llvm.amdgcn.cubetc(float -4.0, float +5.0, float -3.0)
store volatile float %n4p5n3, float* %p
%n5p3n4 = call float @llvm.amdgcn.cubetc(float -5.0, float +3.0, float -4.0)
store volatile float %n5p3n4, float* %p
%n5p4n3 = call float @llvm.amdgcn.cubetc(float -5.0, float +4.0, float -3.0)
store volatile float %n5p4n3, float* %p
%n3n4p5 = call float @llvm.amdgcn.cubetc(float -3.0, float -4.0, float +5.0)
store volatile float %n3n4p5, float* %p
%n3n5p4 = call float @llvm.amdgcn.cubetc(float -3.0, float -5.0, float +4.0)
store volatile float %n3n5p4, float* %p
%n4n3p5 = call float @llvm.amdgcn.cubetc(float -4.0, float -3.0, float +5.0)
store volatile float %n4n3p5, float* %p
%n4n5p3 = call float @llvm.amdgcn.cubetc(float -4.0, float -5.0, float +3.0)
store volatile float %n4n5p3, float* %p
%n5n3p4 = call float @llvm.amdgcn.cubetc(float -5.0, float -3.0, float +4.0)
store volatile float %n5n3p4, float* %p
%n5n4p3 = call float @llvm.amdgcn.cubetc(float -5.0, float -4.0, float +3.0)
store volatile float %n5n4p3, float* %p
%n3n4n5 = call float @llvm.amdgcn.cubetc(float -3.0, float -4.0, float -5.0)
store volatile float %n3n4n5, float* %p
%n3n5n4 = call float @llvm.amdgcn.cubetc(float -3.0, float -5.0, float -4.0)
store volatile float %n3n5n4, float* %p
%n4n3n5 = call float @llvm.amdgcn.cubetc(float -4.0, float -3.0, float -5.0)
store volatile float %n4n3n5, float* %p
%n4n5n3 = call float @llvm.amdgcn.cubetc(float -4.0, float -5.0, float -3.0)
store volatile float %n4n5n3, float* %p
%n5n3n4 = call float @llvm.amdgcn.cubetc(float -5.0, float -3.0, float -4.0)
store volatile float %n5n3n4, float* %p
%n5n4n3 = call float @llvm.amdgcn.cubetc(float -5.0, float -4.0, float -3.0)
store volatile float %n5n4n3, float* %p
ret void
}