Stream: git-wasmtime

Topic: wasmtime / issue #6377 Revert the lea-based lowering for ...


view this post on Zulip Wasmtime GitHub notifications bot (May 15 2023 at 17:45):

github-actions[bot] commented on issue #6377:

Subscribe to Label Action

cc @cfallin, @fitzgen

<details>
This issue or pull request has been labeled: "cranelift", "cranelift:area:x64", "isle"

Thus the following users have been cc'd because of the following labels:

To subscribe or unsubscribe from this label, edit the <code>.github/subscribe-to-label.json</code> configuration file.

Learn more.
</details>

view this post on Zulip Wasmtime GitHub notifications bot (May 17 2023 at 02:23):

alexcrichton commented on issue #6377:

I've done some further investigation of this with @elliottt. We've got a clif function which before this commit took 15 seconds to codegen and after this commit took ~3 seconds. Profiling showed that virtually all time was spent in this loop. We worked together to develop a better means of merging sets there which reduced the compilation time of the test case in question, both before and after this commit, to ~0.2s. It turns out though that this new means isn't actually necessary since the code has already changed on main! This means that relanding this should be a "simple" matter of updating regalloc2 first and then this can re-land.

This bodes well in my mind in that we were pretty surprised that this was the root cause of the regression. Tickling regalloc2 slightly different though makes a lot of sense and it made what was arguably already a pathological case even worse. Fixing the underlying pathology should provide some nice benefits!


Last updated: Dec 23 2024 at 12:05 UTC