[PATCH] D106280: [X86][AVX] scalar_to_vector(load_scalar()) -> load_vector() for fast dereferencable loads

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon created this revision.
RKSimon added reviewers: craig.topper, pengfei, spatel.
Herald added a subscriber: hiraditya.
RKSimon requested review of this revision.
Herald added a project: LLVM.

As reported on PR51075, we fail to make use of dereferencable 128-bit vector loads for float2 loads which were then being widened for float4 operations, preventing a useful load-fold.

We already do a similar fold for insert_subvector patterns of 128-bit loads with 256-bit dereferencable pointers.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D106280

Files:
  llvm/lib/Target/X86/X86ISelLowering.cpp
  llvm/test/CodeGen/X86/load-partial-dot-product.ll
  llvm/test/CodeGen/X86/load-partial.ll
  llvm/test/CodeGen/X86/masked_gather.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D106280.359790.patch
Type: text/x-patch
Size: 9871 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20210719/5d022edf/attachment.bin>