[llvm-dev] [FPEnv] FNEG instruction (original) (raw)

Cameron McInally via llvm-dev llvm-dev at lists.llvm.org
Wed Sep 26 11:21:34 PDT 2018

On Wed, Sep 26, 2018 at 1:24 PM Sanjay Patel <spatel at rotateright.com> wrote: ...

To bring it back to the question of fneg, let me know if this is an accurate summary: 1. fneg would be nice to have for clarity, but it doesn't make optimization in the default LLVM FP environment any easier/better. 2. We will have to do some preliminary work in the IR optimizer to avoid regressions if we add fneg to the IR. 3. We want fneg as a 1st class instruction even though the related fabs/copysign bitstring ops are intrinsics (because fneg is more common than the others?). 4. Adding fneg to IR means we do not need to add a constrained intrinsic for fneg (likewise, there's no need for constrained fabs/copysign because those intrinsics already exist).

For simplicity's sake, I agree with all this. I could nit-pick a bit, but it's not productive to the FNEG conversation.

Thanks for summarizing, Sanjay. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20180926/6392c1ba/attachment.html>

More information about the llvm-dev mailing list