[PATCH] D33728: [X86][SSE] Improve handling of non-temporal aligned loads

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon created this revision.

PR32743 - Avoid folding of non-temporal aligned loads (when supported by the CPU) into instructions which will cause them to become temporal vector loads and pollute the caches.

PR32744 - Split 256-bit vector non-temporal aligned loads on AVX1 targets to keep them non-temporal.

These can be committed separately but are so inter-related I thought it better to get them reviewed together.


Repository:
  rL LLVM

https://reviews.llvm.org/D33728

Files:
  lib/Target/X86/X86ISelLowering.cpp
  lib/Target/X86/X86InstrFragmentsSIMD.td
  test/CodeGen/X86/fast-isel-nontemporal.ll
  test/CodeGen/X86/nontemporal-loads.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D33728.100877.patch
Type: text/x-patch
Size: 37670 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170531/2964d57c/attachment-0001.bin>