-
Notifications
You must be signed in to change notification settings - Fork 5.3k
Cache copies of register sets and data that varies with AVX512 #81868
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
1. Fix availableRegCount to only be stored in LSRA, where it is used. 2. Make copies of rbmAllFloat/rbmFltCalleeTrash to CodeGen and LSRA classes during construction, so access is "closer", in case TP is impacted. (This also makes it possible to define inline accessors in the `emitter` class.)
|
Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch, @kunalspathak Issue Details
|
|
Related #81847 |
tannergooding
left a comment
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.
Nice small wins back from the initial TP regression!
|
No diffs (as expected) Small TP improvements: linux x64 windows x64 |
|
cc @kunalspathak @anthonycanino @dotnet/jit-contrib |
|
Changes look good to me. Do we know the absolute tp diff now? |
kunalspathak
left a comment
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. Nice TP win!
Compared to pre #79544? No, and the improvement is still small compared to that regression. |
emitterclass.)