[all-commits] [llvm/llvm-project] ef38e6: [GlobalISel] introduce MIFlag::NoConvergent

[Sameer Sahasrabuddhe via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: ef38e6d97f482284262172491b6c13b1b9db95b4
      https://github.com/llvm/llvm-project/commit/ef38e6d97f482284262172491b6c13b1b9db95b4
  Author: Sameer Sahasrabuddhe <sameer.sahasrabuddhe at amd.com>
  Date:   2023-08-20 (Sun, 20 Aug 2023)

  Changed paths:
    M llvm/include/llvm/CodeGen/GlobalISel/CallLowering.h
    M llvm/include/llvm/CodeGen/MachineInstr.h
    M llvm/lib/CodeGen/GlobalISel/CallLowering.cpp
    M llvm/lib/CodeGen/MIRParser/MILexer.cpp
    M llvm/lib/CodeGen/MIRParser/MILexer.h
    M llvm/lib/CodeGen/MIRParser/MIParser.cpp
    M llvm/lib/CodeGen/MIRPrinter.cpp
    M llvm/lib/CodeGen/MachineVerifier.cpp
    M llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp
    M llvm/test/CodeGen/AMDGPU/GlobalISel/dereferenceable-declaration.ll
    M llvm/test/CodeGen/AMDGPU/GlobalISel/dropped_debug_info_assert.ll
    M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-assert-align.ll
    M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-abi-attribute-hints.ll
    M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-implicit-args.ll
    M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-non-fixed.ll
    M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-return-values.ll
    M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-sret.ll
    M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call.ll
    M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-indirect-call.ll
    M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-sibling-call.ll
    A llvm/test/CodeGen/MIR/AMDGPU/noconvergent-invalid.mir
    A llvm/test/CodeGen/MIR/AMDGPU/noconvergent.mir

  Log Message:
  -----------
  [GlobalISel] introduce MIFlag::NoConvergent

Some opcodes in MIR are defined to be convergent by the target by setting
IsConvergent in the corresponding TD file. For example, in AMDGPU, the opcodes
G_SI_CALL and G_INTRINSIC* are marked as convergent. But this is too
conservative, since calls to functions that do not execute convergent operations
should not be marked convergent. This information is available in LLVM IR.

The new flag MIFlag::NoConvergent now allows the IR translator to mark an
instruction as not performing any convergent operations. It is relevant only on
occurrences of opcodes that are marked isConvergent in the target.

Differential Revision: https://reviews.llvm.org/D157475