-
Notifications
You must be signed in to change notification settings - Fork 11.7k
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
[AMDGPU] RFC: Add and optionally use GCNIterativeRPTrackers #88797
Closed
Closed
Changes from 6 commits
Commits
Show all changes
9 commits
Select commit
Hold shift + click to select a range
afb45a5
[AMDGPU] NFCI: Use RegionIdx as key of BBLiveInMap
jrbyrnes 92a0531
[AMDGPU] Add GCNIterativeRPTrackers
jrbyrnes 325b39e
[AMDGPU] Introduce getLiveOutMap
jrbyrnes be2ba38
[AMDGPU] Optionally use new trackers
jrbyrnes 45ddcd8
Add note about recede / advance
jrbyrnes 4dcec51
Guard recomputation of liveOuts in SinkRemat stage
jrbyrnes 1c6ff62
Compatability w/ legacy maxOccupancy scheduler + review comment
jrbyrnes b0a75e4
formatting
jrbyrnes dd877ee
Switch to unsigned + fix bug
jrbyrnes File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
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'm not sure what the key is here. I'm also concerned by building maps for liveness info, that's already encoded in the LiveIntervals. Most of this code is also replicating logic from the generic tracker, and all of this is really hard to get correct.
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'm not sure what to do with this comment.
Currently, these maps are keyed on the initial first instructions of the regions. However, if we are going to requery, we need a different key as these first instructions are likely to change. I've used instead RegionIdx -- which tracks the index in the Regions array. This is similarly used to index all these https://github.com/llvm/llvm-project/blob/a5044e6d505deb79f1b00bb39d11096d29b9c910/llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp#L623C3-L623C10 I guess we could restructure as vector.
While this patch does introduce a LiveOutMap, it doesn't introduce the notion of using a Map to cache live-ins -- this already exists in the code. In this patch, these maps optimize compile time by avoiding recalculation of the live regs.
This is mostly copy-paste from existing code in GCNRegPressure, but, nonetheless, we're running a CQE cycle to root out any correctness / performance issues. Per discussion, I think it's better to slightly duplicate logic for time being, and potentially rework the generic tracker in a long term project.
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've tried to address this in #93089
I've left getLiveRegMap unmodified and used the last MI per region as the key to the liveness info (as is similarly done for BBLiveInMap currently). The result is that we do not introduce any new way of building maps for liveness info. The maps are not strictly necessary, and we could recompute them each time we encounter the same region during scheduling, but they do eliminate redundant calculations.