[PATCH] D126040: [InstCombine] Fold a mul with bool value into and
Allen zhong via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Fri May 20 01:21:22 PDT 2022
Allen added inline comments.
================
Comment at: llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp:312
+ match(Op1, m_And(m_Value(Y), m_One())))
+ return BinaryOperator::CreateAnd(Op0, Op1);
+ }
----------------
craig.topper wrote:
> xbolva00 wrote:
> > grandinj wrote:
> > > Surely the more general case is something like:
> > >
> > > if (ComputeMaxSignificantBits(Op0) ==1 ||
> > > ComputeMaxSignificantBits(Op1)==1)
> > >
> > > ?
> > >
> > Yeah, please more general case.
> ComputeMaxSignificantBits(Op0) ==1 means the input is 0 or -1.
>
> You would need computeKnownBits or MaskedValueIsZero.
Thanks very much , I'll try to use the API ComputeMaxSignificantBits.
and there is some failure cases, for example, it 'll turn the -42 into 42.
```
; X * C (when X is a sext boolean) --> X ? -C : 0
define i32 @mul_sext_bool(i1 %x) {
; CHECK-LABEL: @mul_sext_bool(
; CHECK-NEXT: [[M:%.*]] = select i1 [[X:%.*]], i32 -42, i32 0
; CHECK-NEXT: ret i32 [[M]]
;
%s = sext i1 %x to i32
%m = mul i32 %s, 42
ret i32 %m
}
```
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D126040/new/
https://reviews.llvm.org/D126040
More information about the llvm-commits
mailing list