RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-02-25 15:01:07 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

[InstCombine] fix miscompile from multi-use cttz/ctlz transform

Browse Source 
PR45762:
https://bugs.llvm.org/show_bug.cgi?id=45762
This commit is contained in:
Sanjay Patel 2020-05-01 13:51:21 -04:00
parent 53f4697f7c
commit ba016a3eaf
2 changed files with 6 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
lib/Transforms/InstCombine/InstCombineSelect.cpp
View File

@ -883,10 +883,11 @@ static Value *foldSelectCttzCtlz(ICmpInst *ICI, Value *TrueVal, Value *FalseVal,
return SelectArg;
}
// If the ValueOnZero is not the bitwidth, we can at least make use of the
// fact that the cttz/ctlz result will not be used if the input is zero, so
// it's okay to relax it to undef for that case.
if (II->hasOneUse() && !match(II->getArgOperand(1), m_One()))
// The ValueOnZero is not the bitwidth. But if the cttz/ctlz (and optional
// zext/trunc) have one use (ending at the select), the cttz/ctlz result will
// not be used if the input is zero. Relax to 'undef_on_zero' for that case.
if (II->hasOneUse() && SelectArg->hasOneUse() &&
!match(II->getArgOperand(1), m_One()))
II->setArgOperand(1, ConstantInt::getTrue(II->getContext()));
return nullptr;

2
test/Transforms/InstCombine/select-ctlz-to-cttz.ll
View File

@ -220,7 +220,7 @@ define <2 x i32> @select_clz_to_ctz_vec_with_undef(<2 x i32> %a) {
define i4 @PR45762(i3 %x4) {
; CHECK-LABEL: @PR45762(
; CHECK-NEXT: [[T4:%.*]] = call i3 @llvm.cttz.i3(i3 [[X4:%.*]], i1 true), !range !2
; CHECK-NEXT: [[T4:%.*]] = call i3 @llvm.cttz.i3(i3 [[X4:%.*]], i1 false), !range !2
; CHECK-NEXT: [[T7:%.*]] = zext i3 [[T4]] to i4
; CHECK-NEXT: [[ONE_HOT_16:%.*]] = shl i4 1, [[T7]]
; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i3 [[X4]], 0