[PATCH] D110259: [SVE][Analysis] Tune the cost model according to the tune-cpu attribute

[David Sherwood via Phabricator via llvm-commits]

david-arm created this revision.
david-arm added reviewers: sdesmalen, CarolineConcatto, kmclaughlin, dmgreen.
Herald added subscribers: ctetreau, psnobl, hiraditya, kristof.beyls, tschuett.
Herald added a reviewer: efriedma.
david-arm requested review of this revision.
Herald added a project: LLVM.
Herald added a subscriber: llvm-commits.

This patch introduces a new function:

  AArch64Subtarget::getVScaleForTuning

that attempts to return a value for vscale that can be used for tuning
the cost model when using scalable vectors. The function attempts to
determine vscale in the following order:

1. If the user has specified the CPU to tune for we use that, else
2. If the target CPU was specified we use that, else
3. If the vscale_range attribute is present for the function we use the maximum value, else
4. If none of the above assume vscale=16.

For CPUs of type "generic" I have assumed that vscale=4.

Tests added here:

  Analysis/CostModel/AArch64/sve-gather.ll
  Analysis/CostModel/AArch64/sve-scatter.ll


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D110259

Files:
  llvm/lib/Target/AArch64/AArch64Subtarget.cpp
  llvm/lib/Target/AArch64/AArch64Subtarget.h
  llvm/lib/Target/AArch64/AArch64TargetTransformInfo.h
  llvm/test/Analysis/CostModel/AArch64/sve-gather.ll
  llvm/test/Analysis/CostModel/AArch64/sve-scatter.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D110259.374241.patch
Type: text/x-patch
Size: 33965 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20210922/b4e7186b/attachment.bin>