[PATCH] D157123: [FuncSpec] Rework the discardment logic for unprofitable specializations.

[Alexandros Lamprineas via Phabricator via llvm-commits]

labrinea created this revision.
labrinea added reviewers: ChuanqiXu, chill.
Herald added subscribers: hoy, ormris, hiraditya.
Herald added a project: All.
labrinea requested review of this revision.
Herald added a project: LLVM.

Currently we make an arbitrary comparison between codesize and latency
in order to decide whether to keep a specialization or not. Sometimes
the latency savings are biased in favor of loops because of imprecise
block frequencies, therefore this metric contains a lot of noise. This
patch tries to address the problem as follows:

      

- Reject specializations whose codesize savings are less than X% of the original function size.
- Reject specializations whose latency savings are not at least N times greater than the codesize savings.
- Reject specializations whose inlining bonus is not at least N times greater than the original function size.

I am not saying this is super precise, but at least X and N are
configurable, allowing us to tweak the cost model. Moreover, it
lets us prioritize codesize over latency, which is a less noisy
metric.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D157123

Files:
  llvm/include/llvm/Transforms/IPO/FunctionSpecialization.h
  llvm/lib/Transforms/IPO/FunctionSpecialization.cpp
  llvm/unittests/Transforms/IPO/FunctionSpecializationTest.cpp

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D157123.547289.patch
Type: text/x-patch
Size: 11360 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20230804/0ed4c884/attachment.bin>