RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-04-01 07:11:45 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

InstCombine: Avoid introducing poison values when lowering llvm.amdgcn.[us]bfe

Browse Source 
Summary:
When the 3rd argument to these intrinsics is zero, lowering them
to shift instructions produces poison values, since we end up with
shift amounts equal to the number of bits in the shifted value.  This
means we can only lower these intrinsics if we can prove that the
3rd argument is not zero.

Reviewers: arsenm

Reviewed By: arsenm

Subscribers: bnieuwenhuizen, jvesely, wdng, nhaehnle, llvm-commits

Differential Revision: https://reviews.llvm.org/D53739

llvm-svn: 346422
This commit is contained in:
Tom Stellard 2018-11-08 17:57:57 +00:00
parent 628412cc49
commit 0de137231d
2 changed files with 12 additions and 36 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
lib/Transforms/InstCombine/InstCombineCalls.cpp
View File

@ -3479,22 +3479,14 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) {
bool Signed = II->getIntrinsicID() == Intrinsic::amdgcn_sbfe;
// TODO: Also emit sub if only width is constant.
if (!CWidth && COffset && Offset == 0) {
Constant *KSize = ConstantInt::get(COffset->getType(), IntSize);
Value *ShiftVal = Builder.CreateSub(KSize, II->getArgOperand(2));
ShiftVal = Builder.CreateZExt(ShiftVal, II->getType());
Value *Shl = Builder.CreateShl(Src, ShiftVal);
Value *RightShift = Signed ? Builder.CreateAShr(Shl, ShiftVal)
: Builder.CreateLShr(Shl, ShiftVal);
RightShift->takeName(II);
return replaceInstUsesWith(*II, RightShift);
}
if (!CWidth || !COffset)
break;
// The case of Width == 0 is handled above, which makes this tranformation
// safe. If Width == 0, then the ashr and lshr instructions become poison
// value since the shift amount would be equal to the bit size.
assert(Width != 0);
// TODO: This allows folding to undef when the hardware has specific
// behavior?
if (Offset + Width < IntSize) {

30
test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
View File

@ -934,32 +934,23 @@ define i32 @ubfe_offset_33(i32 %src, i32 %width) {
}
; CHECK-LABEL: @ubfe_offset_0(
; CHECK-NEXT: %1 = sub i32 32, %width
; CHECK-NEXT: %2 = lshr i32 -1, %1
; CHECK-NEXT: %bfe = and i32 %2, %src
; CHECK-NEXT: ret i32 %bfe
; CHECK-NEXT: %bfe = call i32 @llvm.amdgcn.ubfe.i32(i32 %src, i32 0, i32 %width)
define i32 @ubfe_offset_0(i32 %src, i32 %width) {
%bfe = call i32 @llvm.amdgcn.ubfe.i32(i32 %src, i32 0, i32 %width)
ret i32 %bfe
}
; CHECK-LABEL: @ubfe_offset_32(
; CHECK-NEXT: %1 = sub i32 32, %width
; CHECK-NEXT: %2 = lshr i32 -1, %1
; CHECK-NEXT: %bfe = and i32 %2, %src
; CHECK-NEXT: ret i32 %bfe
; CHECK-NEXT: %bfe = call i32 @llvm.amdgcn.ubfe.i32(i32 %src, i32 0, i32 %width)
define i32 @ubfe_offset_32(i32 %src, i32 %width) {
%bfe = call i32 @llvm.amdgcn.ubfe.i32(i32 %src, i32 32, i32 %width)
ret i32 %bfe
}
; CHECK-LABEL: @ubfe_offset_31(
; CHECK-NEXT: %1 = sub i32 32, %width
; CHECK-NEXT: %2 = lshr i32 -1, %1
; CHECK-NEXT: %bfe = and i32 %2, %src
; CHECK-NEXT: ret i32 %bfe
; CHECK-NEXT: %bfe = call i32 @llvm.amdgcn.ubfe.i32(i32 %src, i32 31, i32 %width)
define i32 @ubfe_offset_31(i32 %src, i32 %width) {
%bfe = call i32 @llvm.amdgcn.ubfe.i32(i32 %src, i32 32, i32 %width)
%bfe = call i32 @llvm.amdgcn.ubfe.i32(i32 %src, i32 31, i32 %width)
ret i32 %bfe
}
@ -1040,11 +1031,7 @@ define i64 @ubfe_offset_33_width_4_i64(i64 %src) {
}
; CHECK-LABEL: @ubfe_offset_0_i64(
; CHECK-NEXT: %1 = sub i32 64, %width
; CHECK-NEXT: %2 = zext i32 %1 to i64
; CHECK-NEXT: %3 = lshr i64 -1, %2
; CHECK-NEXT: %bfe = and i64 %3, %src
; CHECK-NEXT: ret i64 %bfe
; CHECK-NEXT: %bfe = call i64 @llvm.amdgcn.ubfe.i64(i64 %src, i32 0, i32 %width)
define i64 @ubfe_offset_0_i64(i64 %src, i32 %width) {
%bfe = call i64 @llvm.amdgcn.ubfe.i64(i64 %src, i32 0, i32 %width)
ret i64 %bfe
@ -1066,12 +1053,9 @@ declare i32 @llvm.amdgcn.sbfe.i32(i32, i32, i32) nounwind readnone
declare i64 @llvm.amdgcn.sbfe.i64(i64, i32, i32) nounwind readnone
; CHECK-LABEL: @sbfe_offset_31(
; CHECK-NEXT: %1 = sub i32 32, %width
; CHECK-NEXT: %2 = shl i32 %src, %1
; CHECK-NEXT: %bfe = ashr i32 %2, %1
; CHECK-NEXT: ret i32 %bfe
; CHECK-NEXT: %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %src, i32 31, i32 %width)
define i32 @sbfe_offset_31(i32 %src, i32 %width) {
%bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %src, i32 32, i32 %width)
%bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %src, i32 31, i32 %width)
ret i32 %bfe
}