[all-commits] [llvm/llvm-project] e5822d: [FunctionAttrs] Infer argmemonly .

[Florian Hahn via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: e5822ded562fd233ff7468d88da9e2c4f1a96c70
      https://github.com/llvm/llvm-project/commit/e5822ded562fd233ff7468d88da9e2c4f1a96c70
  Author: Florian Hahn <flo at fhahn.com>
  Date:   2022-03-16 (Wed, 16 Mar 2022)

  Changed paths:
    M llvm/lib/Transforms/IPO/FunctionAttrs.cpp
    M llvm/test/Analysis/TypeBasedAliasAnalysis/functionattrs.ll
    M llvm/test/CodeGen/AMDGPU/inline-attr.ll
    M llvm/test/Transforms/FunctionAttrs/argmemonly.ll
    M llvm/test/Transforms/FunctionAttrs/atomic.ll
    M llvm/test/Transforms/FunctionAttrs/nofree.ll
    M llvm/test/Transforms/FunctionAttrs/nosync.ll
    M llvm/test/Transforms/FunctionAttrs/stats.ll

  Log Message:
  -----------
  [FunctionAttrs] Infer argmemonly .

This patch adds initial argmemonly inference, by checking the underlying
objects of locations returned by MemoryLocation.

I think this should cover most cases, except function calls to other
argmemonly functions.

I'm not sure if there's a reason why we don't infer those yet.

Additional argmemonly can improve codegen in some cases. It also makes
it easier to come up with a C reproducer for 7662d1687b09 (already fixed,
but I'm trying to see if C/C++ fuzzing could help to uncover similar
issues.)

Compile-time impact:
NewPM-O3: +0.01%
NewPM-ReleaseThinLTO: +0.03%
NewPM-ReleaseLTO+g: +0.05%

https://llvm-compile-time-tracker.com/compare.php?from=067c035012fc061ad6378458774ac2df117283c6&to=fe209d4aab5b593bd62d18c0876732ddcca1614d&stat=instructions

Reviewed By: nikic

Differential Revision: https://reviews.llvm.org/D121415