[PATCH] D91919: [SLP] Make SLPVectorizer to use `llvm.masked.scatter` intrinsic

[Anton Afanasyev via Phabricator via llvm-commits]

anton-afanasyev created this revision.
anton-afanasyev added reviewers: ABataev, RKSimon, dtemirbulatov.
Herald added subscribers: llvm-commits, pengfei, hiraditya, kristof.beyls.
Herald added a project: LLVM.
anton-afanasyev requested review of this revision.

This is symmetrical to fcad8d3635c <https://reviews.llvm.org/rGfcad8d3635cff61a2749dcef94c0d51fa1e3e413> commit adding `llvm.masked.gather`.
For the scattered operands of store instructions it makes sense
to use scattered store intrinsic, which can lower to native instruction
for X86/AVX512 and ARM/SVE. This also enables building
vectorization tree with entries containing scattered operands.

Actually NFC now, since stores could be only seed entry, but we
collect only consecutive stores there for now. It will be changed
in future commits.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D91919

Files:
  llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D91919.306861.patch
Type: text/x-patch
Size: 4723 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20201121/c95926af/attachment.bin>