This is an archive of the discontinued LLVM Phabricator instance.

Differential D21746

[AVX512] Zero extend cmp intrinsic return value.
ClosedPublic

Authored by igorb on Jun 27 2016, 3:55 AM.

Details

Reviewers
delena
craig.topper
Commits
rG2c880cf9b131: [AVX512] Zero extend cmp intrinsic return value.
rC274110: [AVX512] Zero extend cmp intrinsic return value.
rL274110: [AVX512] Zero extend cmp intrinsic return value.
Summary

[AVX512] Zero extend cmp intrinsic return value.

Diff Detail

Repository
rL LLVM

Event Timeline

igorb updated this revision to Diff 61953.Jun 27 2016, 3:55 AM
igorb retitled this revision from to [AVX512] Zero extend cmp intrinsic return value..
igorb updated this object.
igorb added reviewers: delena, craig.topper.
igorb set the repository for this revision to rL LLVM.
igorb added a subscriber: cfe-commits.
delena accepted this revision.Jun 27 2016, 9:15 AM
delena edited edge metadata.
This revision is now accepted and ready to land.Jun 27 2016, 9:15 AM
Closed by commit rL274110: [AVX512] Zero extend cmp intrinsic return value. (authored by ibreger). · Explain WhyJun 29 2016, 1:21 AM
This revision was automatically updated to reflect the committed changes.

Revision Contents

PathSize
cfe/
trunk/
lib/
CodeGen/
4 lines
test/
CodeGen/
1 line

Diff 62185

cfe/trunk/lib/CodeGen/CGBuiltin.cpp

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 6,454 Lines • ▼ Show 20 Lines if (!C || !C->isAllOnesValue())
Cmp = CGF.Builder.CreateAnd(Cmp, getMaskVecValue(CGF, Ops.back(), NumElts)); Cmp = CGF.Builder.CreateAnd(Cmp, getMaskVecValue(CGF, Ops.back(), NumElts));
if (NumElts < 8) { if (NumElts < 8) {
uint32_t Indices[8]; uint32_t Indices[8];
for (unsigned i = 0; i != NumElts; ++i) for (unsigned i = 0; i != NumElts; ++i)
Indices[i] = i; Indices[i] = i;
for (unsigned i = NumElts; i != 8; ++i) for (unsigned i = NumElts; i != 8; ++i)
Indices[i] = NumElts; Indices[i] = NumElts;
Cmp = CGF.Builder.CreateShuffleVector(Cmp, UndefValue::get(Cmp->getType()), Cmp = CGF.Builder.CreateShuffleVector(
Indices); Cmp, llvm::Constant::getNullValue(Cmp->getType()), Indices);
} }
return CGF.Builder.CreateBitCast(Cmp, return CGF.Builder.CreateBitCast(Cmp,
IntegerType::get(CGF.getLLVMContext(), IntegerType::get(CGF.getLLVMContext(),
std::max(NumElts, 8U))); std::max(NumElts, 8U)));
} }
Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID, Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
const CallExpr *E) { const CallExpr *E) {
▲ Show 20 Lines • Show All 1,368 Lines • Show Last 20 Lines

cfe/trunk/test/CodeGen/avx512vl-builtins.c

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
// RUN: %clang_cc1 %s -triple=x86_64-apple-darwin -target-feature +avx512f -target-feature +avx512vl -emit-llvm -o - -Werror | FileCheck %s // RUN: %clang_cc1 %s -triple=x86_64-apple-darwin -target-feature +avx512f -target-feature +avx512vl -emit-llvm -o - -Werror | FileCheck %s
// Don't include mm_malloc.h, it's system specific. // Don't include mm_malloc.h, it's system specific.
#define __MM_MALLOC_H #define __MM_MALLOC_H
#include <immintrin.h> #include <immintrin.h>
__mmask8 test_mm_cmpeq_epu32_mask(__m128i __a, __m128i __b) { __mmask8 test_mm_cmpeq_epu32_mask(__m128i __a, __m128i __b) {
// CHECK-LABEL: @test_mm_cmpeq_epu32_mask // CHECK-LABEL: @test_mm_cmpeq_epu32_mask
// CHECK: icmp eq <4 x i32> %{{.*}}, %{{.*}} // CHECK: icmp eq <4 x i32> %{{.*}}, %{{.*}}
// CHECK: shufflevector <4 x i1> %{{.*}}, <4 x i1> zeroinitializer, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 4, i32 4, i32 4>
return (__mmask8)_mm_cmpeq_epu32_mask(__a, __b); return (__mmask8)_mm_cmpeq_epu32_mask(__a, __b);
} }
__mmask8 test_mm_mask_cmpeq_epu32_mask(__mmask8 __u, __m128i __a, __m128i __b) { __mmask8 test_mm_mask_cmpeq_epu32_mask(__mmask8 __u, __m128i __a, __m128i __b) {
// CHECK-LABEL: @test_mm_mask_cmpeq_epu32_mask // CHECK-LABEL: @test_mm_mask_cmpeq_epu32_mask
// CHECK: icmp eq <4 x i32> %{{.*}}, %{{.*}} // CHECK: icmp eq <4 x i32> %{{.*}}, %{{.*}}
// CHECK: and <4 x i1> %{{.*}}, %{{.*}} // CHECK: and <4 x i1> %{{.*}}, %{{.*}}
return (__mmask8)_mm_mask_cmpeq_epu32_mask(__u, __a, __b); return (__mmask8)_mm_mask_cmpeq_epu32_mask(__u, __a, __b);
▲ Show 20 Lines • Show All 6,831 Lines • Show Last 20 Lines