Page MenuHomePhabricator

[NFC][X86] lowerVECTOR_SHUFFLE(): drop FIXME about widening to i128 (YMM half) element type
ClosedPublic

Authored by lebedev.ri on Mon, Jun 7, 8:17 AM.

Details

Summary

As per the discussion in D103818, so far, this does not appear to be worthwhile.

Diff Detail

Event Timeline

lebedev.ri created this revision.Mon, Jun 7, 8:17 AM
lebedev.ri requested review of this revision.Mon, Jun 7, 8:17 AM

Why is this better? vinsertf128 tends to be faster than broadcasts

llvm/test/CodeGen/X86/vector-shuffle-256-v4.ll
1016 ↗(On Diff #350299)

The AVX1 shuffle looks to be much better............

lebedev.ri updated this revision to Diff 350321.Mon, Jun 7, 9:14 AM
lebedev.ri retitled this revision from [X86] lowerVECTOR_SHUFFLE(): allow widening shuffle to have i128 (YMM half) element type to [NFC][X86] lowerVECTOR_SHUFFLE(): drop FIXME about widening to i128 (YMM half) element type.
lebedev.ri edited the summary of this revision. (Show Details)
RKSimon accepted this revision.Mon, Jun 7, 12:35 PM

LGTM

This revision is now accepted and ready to land.Mon, Jun 7, 12:35 PM