[PATCH] D38154: [PassManager] Run global opts after the inliner

[Chandler Carruth via Phabricator via llvm-commits]

chandlerc added a comment.

In https://reviews.llvm.org/D38154#889129, @andjo403 wrote:

> I did a small test in rust with -Os and the compile time do not change with the patch so a check for `OptLevel > 2` is maybe good to have.


Note that "no change" is really quite expected in many cases... I don't think that makes this change inappropriate.

I'm somewhat confused by people's concerns here. This change seems *really* safe:

1. The passes added are among the fastest passes in LLVM.
2. They cover a clear, demonstrable weakness of the rest of the pipeline, deleting dead IR earlier and avoiding expensive work on that IR.

It's of course good to check this stuff, but honestly this isn't the change I would be worried about for compile time...


https://reviews.llvm.org/D38154