[libc-commits] [PATCH] D153823: [libc][math] Fix floating-point test support on x86_64 Apple machines

[Dominic Chen via Phabricator via libc-commits]

ddcc added inline comments.


================
Comment at: libc/src/__support/FPUtil/x86_64/FEnvImpl.h:220
+  uint32_t mxcsr = internal::get_mxcsr();
   // Check both x87 status word and MXCSR.
+  uint16_t status_value = internal::get_status_value_for_except(excepts);
----------------
lntue wrote:
> ddcc wrote:
> > sivachandra wrote:
> > > ddcc wrote:
> > > > Just wanted to point out that this change will affect all x86_64 platforms. But based on the comment, I believe this function is supposed to have been fetching the exception status bits from bot hteh x87 FPU and the SSE unit anyway?
> > > Thanks for catching this! If the tests are passing, go ahead and submit. If you can actually craft a test case to catch regressions, even better.
> > Do you know if the pre-merge checks run the tests at all? I don't have a non-Apple x86_64 machine easily accessible
> I think the pre-merge checks are run on a Linux x86_64 machine, probably in overlay mode.  Post-commit bots will run them in full build mode.
Ok, ran the tests in overlay mode locally on a Linux machine, and all passed.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D153823/new/

https://reviews.llvm.org/D153823