This is an archive of the discontinued LLVM Phabricator instance.

[X86][AVX512] extend support in Scalar conversion
ClosedPublic

Authored by AsafBadouh on Sep 6 2015, 3:34 AM.

Details

Summary

add scalar FP to Int conversion with truncation intrinsics
add scalar conversion FP32 from/to FP64 intrinsics
add rounding mode and SAE mode encoding for these intrinsics

Diff Detail

Repository
rL LLVM

Event Timeline

AsafBadouh updated this revision to Diff 34119.Sep 6 2015, 3:34 AM
AsafBadouh retitled this revision from to [X86][AVX512] extend support in Scalar conversion.
AsafBadouh updated this object.
AsafBadouh added reviewers: delena, igorb.
AsafBadouh added a subscriber: llvm-commits.
delena added inline comments.Sep 8 2015, 12:07 AM
../llvm/lib/Target/X86/X86ISelLowering.cpp
19331 ↗(On Diff #34119)

Please try to avoid adding new nodes. May be you can use the existing staff.

../llvm/lib/Target/X86/X86InstrAVX512.td
4494 ↗(On Diff #34119)

You don't use the "Int" parameter here.

see the upcoming patch

../llvm/lib/Target/X86/X86ISelLowering.cpp
19331 ↗(On Diff #34119)

removed, used VFPROUND instead

../llvm/lib/Target/X86/X86InstrAVX512.td
4494 ↗(On Diff #34119)

fixed and removed

AsafBadouh updated this revision to Diff 34216.Sep 8 2015, 7:55 AM

apply comments

delena edited edge metadata.Sep 19 2015, 11:28 PM

Some minor fixes. You can commit afterwards.

../llvm/lib/Target/X86/X86InstrAVX512.td
4597 ↗(On Diff #34216)

You have two pairs of names avx512_cvt_fp_scalar_rc and avx512_cvt_fp_rc_scalar,

avx512_cvt_fp_scalar_sae and avx512_cvt_fp_sae_scalar, please try to rename

../llvm/lib/Target/X86/X86IntrinsicsInfo.h
562 ↗(On Diff #34216)

The line is too long

This revision was automatically updated to reflect the committed changes.