[all-commits] [llvm/llvm-project] b87e3e: AMDGPU: Don't create potentially dead rcp declarat...

[Matt Arsenault via All-commits]

  Branch: refs/heads/master
  Home:   https://github.com/llvm/llvm-project
  Commit: b87e3e2d0db8b0eb7a8f26525dad74cc7014b85f
      https://github.com/llvm/llvm-project/commit/b87e3e2d0db8b0eb7a8f26525dad74cc7014b85f
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2020-02-11 (Tue, 11 Feb 2020)

  Changed paths:
    M llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp

  Log Message:
  -----------
  AMDGPU: Don't create potentially dead rcp declarations

This will introduce unused declarations if this doesn't reach any of
the paths that will really use it.


  Commit: 92c62582fc546c56b73f78402291337a24acf54e
      https://github.com/llvm/llvm-project/commit/92c62582fc546c56b73f78402291337a24acf54e
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2020-02-11 (Tue, 11 Feb 2020)

  Changed paths:
    M llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp
    M llvm/test/CodeGen/AMDGPU/amdgpu-codegenprepare-fold-binop-select.ll
    M llvm/test/CodeGen/AMDGPU/amdgpu-codegenprepare-idiv.ll
    M llvm/test/CodeGen/AMDGPU/divrem24-assume.ll

  Log Message:
  -----------
  AMDGPU: Directly use rcp intrinsic in idiv expansions

Since natural fdiv lowering is now more conservative even with
denormals disabled, we get a slower expansion from just a plain
1.0/fdiv. Directly emit the rcp intrinsic when using it to implement
integer division to avoid a pointlessly complex sequence.


  Commit: f734ce0488d45c8073892abb6805c9fd9d99fd46
      https://github.com/llvm/llvm-project/commit/f734ce0488d45c8073892abb6805c9fd9d99fd46
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2020-02-11 (Tue, 11 Feb 2020)

  Changed paths:
    M llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
    M llvm/test/CodeGen/AMDGPU/kernel-argument-dag-lowering.ll

  Log Message:
  -----------
  AMDGPU: Fix crash on v3i15 kernel arguments

This was split into 3 i15 arguments. The i15 piece needs to be rounded
to a simple MVT for the memory type.


  Commit: d3a96fc082bef6a15f1ce8aed1849ff4122636dc
      https://github.com/llvm/llvm-project/commit/d3a96fc082bef6a15f1ce8aed1849ff4122636dc
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2020-02-11 (Tue, 11 Feb 2020)

  Changed paths:
    M llvm/test/CodeGen/AMDGPU/amdgpu-codegenprepare-idiv.ll

  Log Message:
  -----------
  AMDGPU: Add baseline tests for CGP div expansion

These cases are harmed by expanding division early in the IR, before
DAGCombiner.


Compare: https://github.com/llvm/llvm-project/compare/bdc3c73454aa...d3a96fc082be