[msan] Support vst1x_{2,3,4} and vst_{2,3,4} with floating-point parameters #100644
+235
−280
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@llvm/pr-subscribers-compiler-rt-sanitizer @llvm/pr-subscribers-llvm-transforms Author: Thurston Dang (thurstond) ChangesCloning the vst_ intrinsics to apply them to the shadows did not work if the arguments were floating-point, since the shadows are integers. This patch changes MSan to create an intrinsic of the correct integer types. Additionally, this patch adds support for vst1x_{2,3,4}; these can be handled similarly to vst_{2,3,4}, since in all cases we are adapting the corresponding intrinsic. This also updates the tests. Patch is 169.10 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/100644.diff 4 Files Affected:
diff --git a/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp b/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
index 910c36faf7e0f..45b3edf2b8f23 100644
--- a/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
+++ b/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
@@ -3873,11 +3873,18 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
setOriginForNaryOp(I);
}
- /// Handle Arm NEON vector store intrinsics (vst{2,3,4}).
+ /// Handle Arm NEON vector store intrinsics (vst{2,3,4} and vst1x_{2,3,4}).
///
/// Arm NEON vector store intrinsics have the output address (pointer) as the
/// last argument, with the initial arguments being the inputs. They return
/// void.
+ ///
+ /// - st4 interleaves the output e.g., st4 (inA, inB, inC, inD, outP) writes
+ /// abcdabcdabcdabcd... into *outP
+ /// - st1_x4 is non-interleaved e.g., st1_x4 (inA, inB, inC, inD, outP)
+ /// writes aaaa...bbbb...cccc...dddd... into *outP
+ /// These instructions can all be instrumented with essentially the same
+ /// MSan logic, simply by applying the corresponding intrinsic to the shadow.
void handleNEONVectorStoreIntrinsic(IntrinsicInst &I) {
IRBuilder<> IRB(&I);
@@ -3892,11 +3899,12 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
if (ClCheckAccessAddress)
insertShadowCheck(Addr, &I);
+ SmallVector<Value *, 8> Shadows;
// Every arg operand, other than the last one, is an input vector
- IntrinsicInst *ShadowI = cast<IntrinsicInst>(I.clone());
for (int i = 0; i < numArgOperands - 1; i++) {
assert(isa<FixedVectorType>(I.getArgOperand(i)->getType()));
- ShadowI->setArgOperand(i, getShadow(&I, i));
+ Value *Shadow = getShadow(&I, i);
+ Shadows.append(1, Shadow);
}
// MSan's GetShadowTy assumes the LHS is the type we want the shadow for
@@ -3914,13 +3922,20 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
cast<FixedVectorType>(I.getArgOperand(0)->getType())->getElementType(),
cast<FixedVectorType>(I.getArgOperand(0)->getType())->getNumElements() *
(numArgOperands - 1));
- Type *ShadowTy = getShadowTy(OutputVectorTy);
- Value *ShadowPtr, *OriginPtr;
+ Type *OutputShadowTy = getShadowTy(OutputVectorTy);
+
+ Value *OutputShadowPtr, *OutputOriginPtr;
// AArch64 NEON does not need alignment (unless OS requires it)
- std::tie(ShadowPtr, OriginPtr) =
- getShadowOriginPtr(Addr, IRB, ShadowTy, Align(1), /*isStore*/ true);
- ShadowI->setArgOperand(numArgOperands - 1, ShadowPtr);
- ShadowI->insertAfter(&I);
+ std::tie(OutputShadowPtr, OutputOriginPtr) = getShadowOriginPtr(
+ Addr, IRB, OutputShadowTy, Align(1), /*isStore*/ true);
+ Shadows.append(1, OutputShadowPtr);
+
+ // CreateIntrinsic will select the correct (integer) type for the
+ // intrinsic; the original instruction I may have either integer- or
+ // float-type inputs.
+ CallInst *CI =
+ IRB.CreateIntrinsic(IRB.getVoidTy(), I.getIntrinsicID(), Shadows);
+ setShadow(&I, CI);
if (MS.TrackOrigins) {
// TODO: if we modelled the vst* instruction more precisely, we could
@@ -3932,7 +3947,8 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
OC.Add(I.getArgOperand(i));
const DataLayout &DL = F.getDataLayout();
- OC.DoneAndStoreOrigin(DL.getTypeStoreSize(OutputVectorTy), OriginPtr);
+ OC.DoneAndStoreOrigin(DL.getTypeStoreSize(OutputVectorTy),
+ OutputOriginPtr);
}
}
@@ -4277,6 +4293,9 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
setOrigin(&I, getCleanOrigin());
break;
+ case Intrinsic::aarch64_neon_st1x2:
+ case Intrinsic::aarch64_neon_st1x3:
+ case Intrinsic::aarch64_neon_st1x4:
case Intrinsic::aarch64_neon_st2:
case Intrinsic::aarch64_neon_st3:
case Intrinsic::aarch64_neon_st4: {
diff --git a/llvm/test/Instrumentation/MemorySanitizer/AArch64/neon_vst.ll b/llvm/test/Instrumentation/MemorySanitizer/AArch64/neon_vst.ll
index df3bfbaf13bbb..421f00fcbc56b 100644
--- a/llvm/test/Instrumentation/MemorySanitizer/AArch64/neon_vst.ll
+++ b/llvm/test/Instrumentation/MemorySanitizer/AArch64/neon_vst.ll
@@ -22,6 +22,7 @@ define void @st2_8b(<8 x i8> %A, <8 x i8> %B, ptr %P) nounwind sanitize_memory {
; CHECK-NEXT: [[TMP4:%.*]] = ptrtoint ptr [[P]] to i64
; CHECK-NEXT: [[TMP5:%.*]] = xor i64 [[TMP4]], 193514046488576
; CHECK-NEXT: [[TMP6:%.*]] = inttoptr i64 [[TMP5]] to ptr
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], ptr [[TMP6]])
; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP7:%.*]], label [[TMP8:%.*]], !prof [[PROF0:![0-9]+]]
; CHECK: 7:
@@ -29,7 +30,6 @@ define void @st2_8b(<8 x i8> %A, <8 x i8> %B, ptr %P) nounwind sanitize_memory {
; CHECK-NEXT: unreachable
; CHECK: 8:
; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> [[A]], <8 x i8> [[B]], ptr [[P]])
-; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], ptr [[TMP6]])
; CHECK-NEXT: ret void
;
call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> %A, <8 x i8> %B, ptr %P)
@@ -47,6 +47,7 @@ define void @st2_8b_undefA(<8 x i8> %A, <8 x i8> %B, ptr %P) nounwind sanitize_m
; CHECK-NEXT: [[TMP3:%.*]] = ptrtoint ptr [[P]] to i64
; CHECK-NEXT: [[TMP4:%.*]] = xor i64 [[TMP3]], 193514046488576
; CHECK-NEXT: [[TMP5:%.*]] = inttoptr i64 [[TMP4]] to ptr
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP2]], ptr [[TMP5]])
; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP6:%.*]], label [[TMP7:%.*]], !prof [[PROF0]]
; CHECK: 6:
@@ -54,7 +55,6 @@ define void @st2_8b_undefA(<8 x i8> %A, <8 x i8> %B, ptr %P) nounwind sanitize_m
; CHECK-NEXT: unreachable
; CHECK: 7:
; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> undef, <8 x i8> [[B]], ptr [[P]])
-; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP2]], ptr [[TMP5]])
; CHECK-NEXT: ret void
;
call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> undef, <8 x i8> %B, ptr %P)
@@ -72,6 +72,7 @@ define void @st2_8b_undefB(<8 x i8> %A, <8 x i8> %B, ptr %P) nounwind sanitize_m
; CHECK-NEXT: [[TMP3:%.*]] = ptrtoint ptr [[P]] to i64
; CHECK-NEXT: [[TMP4:%.*]] = xor i64 [[TMP3]], 193514046488576
; CHECK-NEXT: [[TMP5:%.*]] = inttoptr i64 [[TMP4]] to ptr
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> [[TMP2]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, ptr [[TMP5]])
; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP6:%.*]], label [[TMP7:%.*]], !prof [[PROF0]]
; CHECK: 6:
@@ -79,7 +80,6 @@ define void @st2_8b_undefB(<8 x i8> %A, <8 x i8> %B, ptr %P) nounwind sanitize_m
; CHECK-NEXT: unreachable
; CHECK: 7:
; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> [[A]], <8 x i8> undef, ptr [[P]])
-; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> [[TMP2]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, ptr [[TMP5]])
; CHECK-NEXT: ret void
;
call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> %A, <8 x i8> undef, ptr %P)
@@ -96,6 +96,7 @@ define void @st2_8b_undefAB(<8 x i8> %A, <8 x i8> %B, ptr %P) nounwind sanitize_
; CHECK-NEXT: [[TMP2:%.*]] = ptrtoint ptr [[P]] to i64
; CHECK-NEXT: [[TMP3:%.*]] = xor i64 [[TMP2]], 193514046488576
; CHECK-NEXT: [[TMP4:%.*]] = inttoptr i64 [[TMP3]] to ptr
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, ptr [[TMP4]])
; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP5:%.*]], label [[TMP6:%.*]], !prof [[PROF0]]
; CHECK: 5:
@@ -103,7 +104,6 @@ define void @st2_8b_undefAB(<8 x i8> %A, <8 x i8> %B, ptr %P) nounwind sanitize_
; CHECK-NEXT: unreachable
; CHECK: 6:
; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> undef, <8 x i8> undef, ptr [[P]])
-; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, ptr [[TMP4]])
; CHECK-NEXT: ret void
;
call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> undef, <8 x i8> undef, ptr %P)
@@ -123,6 +123,7 @@ define void @st3_8b(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounwind sani
; CHECK-NEXT: [[TMP5:%.*]] = ptrtoint ptr [[P]] to i64
; CHECK-NEXT: [[TMP6:%.*]] = xor i64 [[TMP5]], 193514046488576
; CHECK-NEXT: [[TMP7:%.*]] = inttoptr i64 [[TMP6]] to ptr
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], ptr [[TMP7]])
; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP8:%.*]], label [[TMP9:%.*]], !prof [[PROF0]]
; CHECK: 8:
@@ -130,7 +131,6 @@ define void @st3_8b(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounwind sani
; CHECK-NEXT: unreachable
; CHECK: 9:
; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> [[A]], <8 x i8> [[B]], <8 x i8> [[C]], ptr [[P]])
-; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], ptr [[TMP7]])
; CHECK-NEXT: ret void
;
call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P)
@@ -149,6 +149,7 @@ define void @st3_8b_undefA(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounwi
; CHECK-NEXT: [[TMP4:%.*]] = ptrtoint ptr [[P]] to i64
; CHECK-NEXT: [[TMP5:%.*]] = xor i64 [[TMP4]], 193514046488576
; CHECK-NEXT: [[TMP6:%.*]] = inttoptr i64 [[TMP5]] to ptr
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP2]], <8 x i8> [[TMP3]], ptr [[TMP6]])
; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP7:%.*]], label [[TMP8:%.*]], !prof [[PROF0]]
; CHECK: 7:
@@ -156,7 +157,6 @@ define void @st3_8b_undefA(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounwi
; CHECK-NEXT: unreachable
; CHECK: 8:
; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> undef, <8 x i8> [[B]], <8 x i8> [[C]], ptr [[P]])
-; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP2]], <8 x i8> [[TMP3]], ptr [[TMP6]])
; CHECK-NEXT: ret void
;
call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> undef, <8 x i8> %B, <8 x i8> %C, ptr %P)
@@ -175,6 +175,7 @@ define void @st3_8b_undefB(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounwi
; CHECK-NEXT: [[TMP4:%.*]] = ptrtoint ptr [[P]] to i64
; CHECK-NEXT: [[TMP5:%.*]] = xor i64 [[TMP4]], 193514046488576
; CHECK-NEXT: [[TMP6:%.*]] = inttoptr i64 [[TMP5]] to ptr
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> [[TMP2]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP3]], ptr [[TMP6]])
; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP7:%.*]], label [[TMP8:%.*]], !prof [[PROF0]]
; CHECK: 7:
@@ -182,7 +183,6 @@ define void @st3_8b_undefB(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounwi
; CHECK-NEXT: unreachable
; CHECK: 8:
; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> [[A]], <8 x i8> undef, <8 x i8> [[C]], ptr [[P]])
-; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> [[TMP2]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP3]], ptr [[TMP6]])
; CHECK-NEXT: ret void
;
call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> %A, <8 x i8> undef, <8 x i8> %C, ptr %P)
@@ -201,6 +201,7 @@ define void @st3_8b_undefC(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounwi
; CHECK-NEXT: [[TMP4:%.*]] = ptrtoint ptr [[P]] to i64
; CHECK-NEXT: [[TMP5:%.*]] = xor i64 [[TMP4]], 193514046488576
; CHECK-NEXT: [[TMP6:%.*]] = inttoptr i64 [[TMP5]] to ptr
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, ptr [[TMP6]])
; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP7:%.*]], label [[TMP8:%.*]], !prof [[PROF0]]
; CHECK: 7:
@@ -208,7 +209,6 @@ define void @st3_8b_undefC(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounwi
; CHECK-NEXT: unreachable
; CHECK: 8:
; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> [[A]], <8 x i8> [[B]], <8 x i8> undef, ptr [[P]])
-; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, ptr [[TMP6]])
; CHECK-NEXT: ret void
;
call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> %A, <8 x i8> %B, <8 x i8> undef, ptr %P)
@@ -226,6 +226,7 @@ define void @st3_8b_undefAB(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounw
; CHECK-NEXT: [[TMP3:%.*]] = ptrtoint ptr [[P]] to i64
; CHECK-NEXT: [[TMP4:%.*]] = xor i64 [[TMP3]], 193514046488576
; CHECK-NEXT: [[TMP5:%.*]] = inttoptr i64 [[TMP4]] to ptr
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP2]], ptr [[TMP5]])
; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP6:%.*]], label [[TMP7:%.*]], !prof [[PROF0]]
; CHECK: 6:
@@ -233,7 +234,6 @@ define void @st3_8b_undefAB(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounw
; CHECK-NEXT: unreachable
; CHECK: 7:
; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> undef, <8 x i8> undef, <8 x i8> [[C]], ptr [[P]])
-; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP2]], ptr [[TMP5]])
; CHECK-NEXT: ret void
;
call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> undef, <8 x i8> undef, <8 x i8> %C, ptr %P)
@@ -251,6 +251,7 @@ define void @st3_8b_undefAC(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounw
; CHECK-NEXT: [[TMP3:%.*]] = ptrtoint ptr [[P]] to i64
; CHECK-NEXT: [[TMP4:%.*]] = xor i64 [[TMP3]], 193514046488576
; CHECK-NEXT: [[TMP5:%.*]] = inttoptr i64 [[TMP4]] to ptr
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP2]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, ptr [[TMP5]])
; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP6:%.*]], label [[TMP7:%.*]], !prof [[PROF0]]
; CHECK: 6:
@@ -258,7 +259,6 @@ define void @st3_8b_undefAC(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounw
; CHECK-NEXT: unreachable
; CHECK: 7:
; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> undef, <8 x i8> [[B]], <8 x i8> undef, ptr [[P]])
-; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP2]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, ptr [[TMP5]])
; CHECK-NEXT: ret void
;
call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> undef, <8 x i8> %B, <8 x i8> undef, ptr %P)
@@ -276,6 +276,7 @@ define void @st3_8b_undefBC(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounw
; CHECK-NEXT: [[TMP3:%.*]] = ptrtoint ptr [[P]] to i64
; CHECK-NEXT: [[TMP4:%.*]] = xor i64 [[TMP3]], 193514046488576
; CHECK-NEXT: [[TMP5:%.*]] = inttoptr i64 [[TMP4]] to ptr
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> [[TMP2]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, ptr [[TMP5]])
; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP6:%.*]], label [[TMP7:%.*]], !prof [[PROF0]]
; CHECK: 6:
@@ -283,7 +284,6 @@ define void @st3_8b_undefBC(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounw
; CHECK-NEXT: unreachable
; CHECK: 7:
; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> [[A]], <8 x i8> undef, <8 x i8> undef, ptr [[P]])
-; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> [[TMP2]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, ptr [[TMP5]])
; CHECK-NEXT: ret void
;
call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> %A, <8 x i8> undef, <8 x i8> undef, ptr %P)
@@ -300,6 +300,7 @@ define void @st3_8b_undefABC(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) noun
; CHECK-NEXT: [[TMP2:%.*]] = ptrtoint ptr [[P]] to i64
; CHECK-NEXT: [[TMP3:%.*]] = xor i64 [[TMP2]], 193514046488576
; CHECK-NEXT: [[TMP4:%.*]] = inttoptr i64 [[TMP3]] to ptr
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, ptr [[TMP4]])
; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP5:%.*]], label [[TMP6:%.*]], !prof [[PROF0]]
; CHECK: 5:
@@ -307,7 +308,6 @@ define void @st3_8b_undefABC(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) noun
; CHECK-NEXT: unreachable
; CHECK: 6:
; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> undef, <8 x i8> undef, <8 x i8> undef, ptr [[P]])
-; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, ptr [[TMP4]])
; CHECK-NEXT: ret void
;
call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> undef, <8 x i8> undef, <8 x i8> undef, ptr %P)
@@ -328,6 +328,7 @@ define void @st4_8b(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P)
; CHECK-NEXT: [[TMP6:%.*]] = ptrtoint ptr [[P]] to i64
; CHECK-NEXT: [[TMP7:%.*]] = xor i64 [[TMP6]], 193514046488576
; CHECK-NEXT: [[TMP8:%.*]] = inttoptr i64 [[TMP7]] to ptr
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], ptr [[TMP8]])
; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP9:%.*]], label [[TMP10:%.*]], !prof [[PROF0]]
; CHECK: 9:
@@ -335,7 +336,6 @@ define void @st4_8b(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P)
; CHECK-NEXT: unreachable
; CHECK: 10:
; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> [[A]], <8 x i8> [[B]], <8 x i8> [[C]], <8 x i8> [[D]], ptr [[P]])
-; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], ptr [[TMP8]])
; CHECK-NEXT: ret void
;
call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P)
@@ -355,6 +355,7 @@ define void @st4_8b_undefA(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, p
; CHECK-NEXT: [[TMP5:%.*]] = ptrtoint ptr [[P]] to i64
; CHECK-NEXT: [[TMP6:%.*]] = xor i64 [[TMP5]], 193514046488576
; CHECK-NEXT: [[TMP7:%.*]] = inttoptr i64 [[TMP6]] to ptr
+; CHECK-NEXT: call void @llvm...
[truncated]
|
…meters
Cloning the vst_ functions to apply them to the shadows did not work if the arguments were floating-point, since the shadows are integers. This patch changes MSan to create an intrinsic of the correct integer types.
Additionally, this patch adds support for vst1x_{2,3,4}; these can be handled similarly to vst_{2,3,4}, since in all cases we are cloning/adapting the corresponding function.
This also updates and enables the test introduced in llvm#100189
vitalybuka
force-pushed
the
msan_vst_float
branch
from
8c7cb2c to
c2261b6
Compare
vitalybuka
approved these changes
Jul 29, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Cloning the vst_ intrinsics to apply them to the shadows did not work if the arguments were floating-point, since the shadows are integers. This patch changes MSan to create an intrinsic of the correct integer types.
Additionally, this patch adds support for vst1x_{2,3,4}; these can be handled similarly to vst_{2,3,4}, since in all cases we are adapting the corresponding intrinsic.
This also updates the tests.