[all-commits] [llvm/llvm-project] fe0750: SeparateConstOffsetFromGEP: Reorder run lines

[Matt Arsenault via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: fe0750b97943070c4dc652afa3628aee0b0d0ebb
      https://github.com/llvm/llvm-project/commit/fe0750b97943070c4dc652afa3628aee0b0d0ebb
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2023-06-26 (Mon, 26 Jun 2023)

  Changed paths:
    M llvm/test/Transforms/SeparateConstOffsetFromGEP/NVPTX/split-gep-and-gvn.ll

  Log Message:
  -----------
  SeparateConstOffsetFromGEP: Reorder run lines

Testing codegen in test/Transforms is questionable to begin with, but
it's more reasonable to see failures on the IR half before ISA checks.


  Commit: e9ab7ff73aea954669e9c52aa3917cee6bb6fb83
      https://github.com/llvm/llvm-project/commit/e9ab7ff73aea954669e9c52aa3917cee6bb6fb83
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2023-06-26 (Mon, 26 Jun 2023)

  Changed paths:
    A llvm/test/Transforms/SeparateConstOffsetFromGEP/AMDGPU/split-gep-and-gvn.ll

  Log Message:
  -----------
  SeparateConstOffsetFromGEP: Copy a test to AMDGPU


  Commit: 6882d9adb4094b3122a49d287b0d944405bbc763
      https://github.com/llvm/llvm-project/commit/6882d9adb4094b3122a49d287b0d944405bbc763
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2023-06-26 (Mon, 26 Jun 2023)

  Changed paths:
    A llvm/test/Transforms/SeparateConstOffsetFromGEP/AMDGPU/reunite-exts-source-types.ll
    A llvm/test/Transforms/SeparateConstOffsetFromGEP/AMDGPU/reunite-exts.ll

  Log Message:
  -----------
  SeparateConstOffsetForGEP: Fill out some missing test coverage

Try to test several untested paths.
 - Test the extension source type check
 - Test the programUndefinedIfPoison check
 - Test the add/sub with commuted operands
 - Test with vectors
 - Test multiple uses
 - Try to break operand map mismatches
 - Add some preparatory tests for zext+nuw support.


  Commit: 574f2bac18803a85020efe2663c53988892ba1ec
      https://github.com/llvm/llvm-project/commit/574f2bac18803a85020efe2663c53988892ba1ec
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2023-06-26 (Mon, 26 Jun 2023)

  Changed paths:
    M llvm/lib/Transforms/Scalar/SeparateConstOffsetFromGEP.cpp

  Log Message:
  -----------
  SeparateConstOffsetForGEP: Remove some typed pointer code


  Commit: f2596b754c2f286cc4df2feff27b3e14a88d4f7b
      https://github.com/llvm/llvm-project/commit/f2596b754c2f286cc4df2feff27b3e14a88d4f7b
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2023-06-26 (Mon, 26 Jun 2023)

  Changed paths:
    M llvm/lib/Transforms/Scalar/SeparateConstOffsetFromGEP.cpp
    M llvm/test/CodeGen/AMDGPU/llc-pipeline.ll
    M llvm/test/CodeGen/PowerPC/O3-pipeline.ll
    M llvm/test/Transforms/SeparateConstOffsetFromGEP/AMDGPU/reunite-exts.ll

  Log Message:
  -----------
  SeparateConstOffsetFromGEP: Don't use SCEV

This was only using the SCEV expressions as a map key, which we can do
just as well with the value pointers. This also allows it to handle
vectors.


Compare: https://github.com/llvm/llvm-project/compare/249bd9eab0aa...f2596b754c2f