hkmatsumoto added inline comments.
================ Comment at: llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp:915 +/// Fold (icmp eq ctpop(X) 1) | (icmp eq X 0) into (icmp ult ctpop(X) 2) and +/// fold (icmp ne ctpop(X) 1) & (icmp ne X 0) into (icmp uge ctpop(X) 2). +static Value *foldIsPowerOf2OrZero(ICmpInst *Cmp0, ICmpInst *Cmp1, bool IsAnd, ---------------- spatel wrote: > Why create ">= 2" instead of "> 1" directly? > > I don't think it makes the transform or code any clearer with ">= 2", and we > will always canonicalize to the other form, so I would prefer to go directly > to the final result for efficiency. You're right. I thought it is fine to let another pass do further transformation but in retrospect, folding >= 2 to > 1 is so trivial that we should do it here. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D122077/new/ https://reviews.llvm.org/D122077 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
