[all-commits] [llvm/llvm-project] fced87: [AMDGPU] Fix regression with vectorization limiting

[Stanislav Mekhanoshin via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: fced87d457d39d94cf7b396d8200dcb4e171003d
      https://github.com/llvm/llvm-project/commit/fced87d457d39d94cf7b396d8200dcb4e171003d
  Author: Stanislav Mekhanoshin <Stanislav.Mekhanoshin at amd.com>
  Date:   2022-04-08 (Fri, 08 Apr 2022)

  Changed paths:
    M llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
    M llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h
    A llvm/test/Transforms/LoopVectorize/AMDGPU/packed-fp32.ll
    M llvm/test/Transforms/LoopVectorize/AMDGPU/packed-math.ll

  Log Message:
  -----------
  [AMDGPU] Fix regression with vectorization limiting

D67148 has removed TTI::getNumberOfRegisters(bool Vector) and
started to call TTI::getNumberOfRegisters(unsigned ClassID) from
the LoopVectorize. This has resulted in an unrestricted vectorization
on AMDGPU blowing up register pressure.

Differential Revision: https://reviews.llvm.org/D122850