[PATCH] D68651: [InstCombine] Signed saturation patterns

[Dave Green via Phabricator via llvm-commits]

dmgreen added a comment.

Hello, Looks like noise to me: http://lnt.llvm.org/db_default/v4/nts/graph?plot.0=1344.1605916.3&highlight_run=130142

I don't think that benchmark contains anything that would hit this code. If you compare gcc-loops in these two runs, the hashes of the binaries are the same:
https://lnt.llvm.org/api/db_default/v4/nts/runs/130141
https://lnt.llvm.org/api/db_default/v4/nts/runs/130142
Why the first was repeatedly 1.3 and the later repeatedly 1.4, i'm not sure, but it looks like it's been bouncing around like that for a while.

(Also I feel that the hashes not changing should really appear in the run results somewhere! It's useful to filter out the noise)


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D68651/new/

https://reviews.llvm.org/D68651