[InstCombine] fold mul of zext/sext bools to 'and'
Similar to rG40fcc42:
The base case only worked because we were relying on a
poison-unsafe select transform; if that is fixed, we
would regress on patterns like this.
The extra use tests show that the select transform can't
be applied consistently. So it may be a regression to have
an extra instruction on 1 test, but that result was not
created safely and does not happen reliably.
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp b/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
index 2965103..c6233a6 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
@@ -376,6 +376,16 @@
Value *And = Builder.CreateAnd(X, Y, "mulbool");
return CastInst::Create(Instruction::ZExt, And, I.getType());
}
+ // (sext bool X) * (zext bool Y) --> sext (and X, Y)
+ // (zext bool X) * (sext bool Y) --> sext (and X, Y)
+ // Note: -1 * 1 == 1 * -1 == -1
+ if (((match(Op0, m_SExt(m_Value(X))) && match(Op1, m_ZExt(m_Value(Y)))) ||
+ (match(Op0, m_ZExt(m_Value(X))) && match(Op1, m_SExt(m_Value(Y))))) &&
+ X->getType()->isIntOrIntVectorTy(1) && X->getType() == Y->getType() &&
+ (Op0->hasOneUse() || Op1->hasOneUse())) {
+ Value *And = Builder.CreateAnd(X, Y, "mulbool");
+ return CastInst::Create(Instruction::SExt, And, I.getType());
+ }
// (bool X) * Y --> X ? Y : 0
// Y * (bool X) --> X ? Y : 0