-
Notifications
You must be signed in to change notification settings - Fork 5k
JIT: Align newly recognized loops #96756
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
Also remove quirk when checking for inner-most loops.
Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch Issue DetailsAlso remove quirk when checking for inner-most loops.
|
// TODO-Quirk: Remove. When removing we will likely need to add some | ||
// form of "lexicality" heuristic here: only align loops whose blocks | ||
// are fairly tightly packed together physically. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I decided not to look at this here; in particular because we have a lot of work in flight that is severely changing the flow graph, so it's hard to get a good feeling for how this heuristic should eventually look.
cc @dotnet/jit-contrib PTAL @BruceForstall Diffs from a run with alignment enabled. Aggregated stats for our win-x64 SPMI collections: -Alignment candidates: 35898
+Alignment candidates: 38075
-Aligned: 13523
+Aligned: 14193 6.1% more candidates, 5.0% more aligned loops |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
superpmi-replay failures are the timeouts we see every once in a while. The linker tests had a timeout in the macOS job. Other failures are known. |
Also remove quirk when checking for inner-most loops.
Also remove quirk when checking for inner-most loops.