This is an archive of the discontinued LLVM Phabricator instance.

[PowerPC] Exclude frexp(long double) on linux
ClosedPublic

Authored by lei on Aug 22 2023, 1:53 PM.

Details

Summary

PowerPC on linux currently don't have support for lowering long double for
frexp(). Removing the tests until implementation is provided.

Diff Detail

Event Timeline

lei created this revision.Aug 22 2023, 1:53 PM
Herald added a project: Restricted Project. · View Herald TranscriptAug 22 2023, 1:53 PM
lei requested review of this revision.Aug 22 2023, 1:53 PM
Herald added a project: Restricted Project. · View Herald TranscriptAug 22 2023, 1:53 PM
Herald added a reviewer: Restricted Project. · View Herald Transcript
Mordante added inline comments.
libcxx/test/libcxx/numerics/c.math/constexpr-cxx23-clang.pass.cpp
61

I prefer an #if ! defined(__LONG_DOUBLE_IBM128__) here instead of moving this line. That keeps the order as expected.

lei updated this revision to Diff 552874.Aug 23 2023, 1:48 PM

Keep original order

lei marked an inline comment as done.Aug 23 2023, 1:48 PM
Mordante accepted this revision.Aug 25 2023, 9:32 AM

Thanks, LGTM from libc++'s perspective!

I don't know whether the used definition __LONG_DOUBLE_IBM128__ is correct, can one of the PowerPC experts validate that? Please wait for an approval of one of the PowerPC experts before landing.

This revision is now accepted and ready to land.Aug 25 2023, 9:32 AM
amyk accepted this revision.Aug 25 2023, 10:07 AM

I think this LGTM. __LONG_DOUBLE_IBM128__ is what we define (on Linux) when we're dealing with the ppcf128 type.

This revision was automatically updated to reflect the committed changes.