[PATCH] D69870: [builtins] Implement rounding mode support for i386/x86_64

[Craig Topper via Phabricator via llvm-commits]

craig.topper added a comment.

In D69870#1744853 <https://reviews.llvm.org/D69870#1744853>, @kongyi wrote:

> In D69870#1744626 <https://reviews.llvm.org/D69870#1744626>, @craig.topper wrote:
>
> > Would it be better to just do a dummy inexact division operation to trigger than inexact exception instead of manipulating the environment?
>
>
> I assumed this would have better performance than dummy inexact division operation. Do you want me to verify it?




In D69870#1744853 <https://reviews.llvm.org/D69870#1744853>, @kongyi wrote:

> In D69870#1744626 <https://reviews.llvm.org/D69870#1744626>, @craig.topper wrote:
>
> > Would it be better to just do a dummy inexact division operation to trigger than inexact exception instead of manipulating the environment?
>
>
> I assumed this would have better performance than dummy inexact division operation. Do you want me to verify it?


I can't imagine fldenv/fstenv are very fast instructions. I wouldn't be surprised if they serialized the core and prevented speculation past them.


Repository:
  rCRT Compiler Runtime

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D69870/new/

https://reviews.llvm.org/D69870