[PATCH] D49474: [X86][SSE] Canonicalize scalar fp arithmetic shuffle patterns

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon created this revision.
RKSimon added reviewers: craig.topper, dyung, spatel, andreadb, gbedwell.

As discussed on PR38197, this canonicalizes MOVS*(N0, OP(N0, N1)) --> MOVS*(N0, SCALAR_TO_VECTOR(OP(N0[0], N1[0])))

This returns the scalar-fp codegen lost by https://reviews.llvm.org/rL336971.

Additionally it handles the OP(N1, N0)) case for commutable (FADD/FMUL) ops.

@craig.topper - This interferes with https://reviews.llvm.org/D49280 but I've tried to make the merge as simple as possible.


Repository:
  rL LLVM

https://reviews.llvm.org/D49474

Files:
  lib/Target/X86/X86ISelLowering.cpp
  test/CodeGen/X86/sse-scalar-fp-arith.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D49474.156048.patch
Type: text/x-patch
Size: 40622 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180718/bb80b27b/attachment.bin>