[all-commits] [llvm/llvm-project] 392190: CodeGen: Remove UsesMSVCFloatingPoint from Machine...

Matt Arsenault via All-commits all-commits at lists.llvm.org
Fri Jul 26 00:27:33 PDT 2024


  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 3921900198653d205753b2d85dc31ca5bc009fdd
      https://github.com/llvm/llvm-project/commit/3921900198653d205753b2d85dc31ca5bc009fdd
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2024-07-26 (Fri, 26 Jul 2024)

  Changed paths:
    M llvm/include/llvm/CodeGen/MachineModuleInfo.h
    M llvm/lib/CodeGen/MachineModuleInfo.cpp
    M llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
    M llvm/lib/Target/X86/X86AsmPrinter.cpp

  Log Message:
  -----------
  CodeGen: Remove UsesMSVCFloatingPoint from MachineModuleInfo (#100368)

This is only used by x86 and only used in the AsmPrinter module pass. I
think implementing this by looking at the underlying IR types instead
of the selected instructions is a pretty horrifying implementation,
but it's still available in the AsmPrinter.

This is https://reviews.llvm.org/D123933 resurrected.

I still don't know what the point of emitting _fltused is, but this
approach of looking at the IR types probably isn't the right way to
do this in the first place. If the intent is report any FP instructions,
this will miss any implicitly introduced ones during codegen. Also don't
know why just unconditionally emitting it isn't an option.

The last review mentioned the ARMs might want to emit this, but I'm
not going to go fix that. If someone wants to emit this on ARM, they
can move this to a common helper or analysis somewhere.



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications


More information about the All-commits mailing list