RPCS3/llvm
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm.git synced 2024-12-28 07:05:03 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

[msan] Handle x86 vector pack intrinsics.

Browse Source 
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210020 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Evgeniy Stepanov 2014-06-02 12:31:44 +00:00
parent ef8a99c807
commit dd7c5bb730
2 changed files with 74 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

36
lib/Transforms/Instrumentation/MemorySanitizer.cpp
View File

@ -1944,6 +1944,28 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
setOriginForNaryOp(I);
}
// \brief Instrument vector shift instrinsic.
//
// This function instruments intrinsics like x86_mmx_packsswb, that
// packs elements of 2 input vectors into half as much bits with saturation.
// Shadow is propagated with the same intrinsic applied to
// sext(Sa != zeroinitializer), sext(Sb != zeroinitializer).
void handleVectorPackIntrinsic(IntrinsicInst &I) {
assert(I.getNumArgOperands() == 2);
IRBuilder<> IRB(&I);
Value *S1 = getShadow(&I, 0);
Value *S2 = getShadow(&I, 1);
Type *T = S1->getType();
Value *S1_ext = IRB.CreateSExt(
IRB.CreateICmpNE(S1, llvm::Constant::getNullValue(T)), T);
Value *S2_ext = IRB.CreateSExt(
IRB.CreateICmpNE(S2, llvm::Constant::getNullValue(T)), T);
Value *S = IRB.CreateCall2(I.getCalledValue(), S1_ext, S2_ext,
"_msprop_vector_pack");
setShadow(&I, S);
setOriginForNaryOp(I);
}
void visitIntrinsicInst(IntrinsicInst &I) {
switch (I.getIntrinsicID()) {
case llvm::Intrinsic::bswap:
@ -2060,6 +2082,20 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
// case llvm::Intrinsic::x86_sse2_psll_dq_bs:
// case llvm::Intrinsic::x86_sse2_psrl_dq_bs:
case llvm::Intrinsic::x86_sse2_packsswb_128:
case llvm::Intrinsic::x86_sse2_packssdw_128:
case llvm::Intrinsic::x86_sse2_packuswb_128:
case llvm::Intrinsic::x86_sse41_packusdw:
case llvm::Intrinsic::x86_avx2_packsswb:
case llvm::Intrinsic::x86_avx2_packssdw:
case llvm::Intrinsic::x86_avx2_packuswb:
case llvm::Intrinsic::x86_avx2_packusdw:
case llvm::Intrinsic::x86_mmx_packsswb:
case llvm::Intrinsic::x86_mmx_packssdw:
case llvm::Intrinsic::x86_mmx_packuswb:
handleVectorPackIntrinsic(I);
break;
default:
if (!handleUnknownIntrinsic(I))
visitInstruction(I);

38
test/Instrumentation/MemorySanitizer/vector_pack.ll Normal file
View File

@ -0,0 +1,38 @@
; RUN: opt < %s -msan -msan-check-access-address=0 -S | FileCheck %s
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-linux-gnu"
declare <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32>, <4 x i32>) nounwind readnone
declare <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %a, <16 x i16> %b) nounwind readnone
define <8 x i16> @Test_packssdw_128(<4 x i32> %a, <4 x i32> %b) sanitize_memory {
entry:
%c = tail call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b) nounwind
ret <8 x i16> %c
}
; CHECK-LABEL: @Test_packssdw_128(
; CHECK-DAG: icmp ne <4 x i32> {{.*}}, zeroinitializer
; CHECK-DAG: sext <4 x i1> {{.*}} to <4 x i32>
; CHECK-DAG: icmp ne <4 x i32> {{.*}}, zeroinitializer
; CHECK-DAG: sext <4 x i1> {{.*}} to <4 x i32>
; CHECK-DAG: call <8 x i16> @llvm.x86.sse2.packssdw.128(
; CHECK-DAG: call <8 x i16> @llvm.x86.sse2.packssdw.128(
; CHECK: ret <8 x i16>
define <32 x i8> @Test_avx_packuswb(<16 x i16> %a, <16 x i16> %b) sanitize_memory {
entry:
%c = tail call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %a, <16 x i16> %b) nounwind
ret <32 x i8> %c
}
; CHECK-LABEL: @Test_avx_packuswb(
; CHECK-DAG: icmp ne <16 x i16> {{.*}}, zeroinitializer
; CHECK-DAG: sext <16 x i1> {{.*}} to <16 x i16>
; CHECK-DAG: icmp ne <16 x i16> {{.*}}, zeroinitializer
; CHECK-DAG: sext <16 x i1> {{.*}} to <16 x i16>
; CHECK-DAG: call <32 x i8> @llvm.x86.avx2.packuswb(
; CHECK-DAG: call <32 x i8> @llvm.x86.avx2.packuswb(
; CHECK: ret <32 x i8>