[all-commits] [llvm/llvm-project] 502b5e: [MemProf] Dump inline call stacks as optimization ...

[Kazu Hirata via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 502b5e0bea1282006b7ba74b7e4c2df3eea4dd12
      https://github.com/llvm/llvm-project/commit/502b5e0bea1282006b7ba74b7e4c2df3eea4dd12
  Author: Kazu Hirata <kazu at google.com>
  Date:   2026-03-27 (Fri, 27 Mar 2026)

  Changed paths:
    M llvm/lib/Transforms/Instrumentation/MemProfUse.cpp
    A llvm/test/Transforms/PGOProfile/memprof-inline-call-stacks.ll

  Log Message:
  -----------
  [MemProf] Dump inline call stacks as optimization remarks (#188678)

This patch teaches the MemProf matching pass to dump inline call
stacks as analysis remarks like so:

frame: 704e4117e6a62739 main:10:5
frame: 273929e54b9f1234 foo:2:12
inline call stack: 704e4117e6a62739,273929e54b9f1234

The output consists of two types of remarks:

- "frame": Acts as a dictionary mapping a unique MD5-based FrameID
  to source information (function name, line offset, and column).

- "inline call stack": Provides the full call stack for a call site
  as a sequence of FrameIDs.

Both types of remarks are deduplicated to reduce the output size.

This patch is intended to be a debugging aid.



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