[PATCH] D120584: [NewPM][Inliner] Mark inlined calls to functions in same SCC as callee as noinline

[Mircea Trofin via Phabricator via llvm-commits]

mtrofin added a comment.

In D120584#3346280 <https://reviews.llvm.org/D120584#3346280>, @aeubanks wrote:

> will fix the two failing tests if we like this patch
>
> this does cause fewer instances of inlining on llvm-test-suite
>
> I did some performance testing on some internal benchmarks, seeing mixed results

IIUC, we sometimes want to allow even recursive function inlining, up to a point; and same for mutually-recursive ones. For instance, if A<->B, both A and B may be worth inlining in B and A, respectively, for their respective base cases.

Since there are regressions we've observed, sounds like we can isolate 1-2 of those and better understand the problem?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D120584/new/

https://reviews.llvm.org/D120584