Repository navigation
Reland xls scheduler refactor. Fix issues causing regressions on some targets. - #4324
Merged
Merged
Conversation
copybara-service
Bot
force-pushed
the
test_920305238
branch
from
May 28, 2026 17:23
fe20a33 to
35ecdd2
Compare
… targets. A mismatch between how the proto considers the worst-case-throughput flag to be unset and the way the SchedulingOptions struct does led to the initial version of this incorrectly believing the WCT was unspecified in some scenarios when it was fully set. Original Description: Refactor XLS schedulers to use a common interface and allow for selecting scheduler for minimization workloads. This change introduces a `Scheduler` abstract base class, and refactors `ASAP`, `MinCut`, and `SDC` scheduling algorithms to implement this interface. A new `RandomScheduler` is also added. The `run_pipeline_schedule` utility is updated to use this new interface. Additionally, a new option and flag `--find_bounds_strategy` are added, allowing a different scheduler (defaulting to the selected scheduling algorithm usually SDC) to be used for finding minimum clock period and worst-case throughput bounds, which can be faster than using the main scheduling strategy. Error handling in `run_pipeline_schedule` is also improved. The 'random' scheduler is also updated to make use of the ASAP scheduler bounds to do a much better job at consistently choosing allowable schedules. The previous run_pipeline_scheduleing had a number of subtle bugs around how its configuration was interpreted that have all been fixed up, requiring some changes to tests which relied on the previous behavior. Testing is improved to check schedulers other than SDC. Many tests needed to be tweaked to account for the different ways these schedulers work. PiperOrigin-RevId: 922893892
copybara-service
Bot
force-pushed
the
test_920305238
branch
from
May 28, 2026 18:52
35ecdd2 to
5e7f3fe
Compare
This branch was previously deployed
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Reland xls scheduler refactor. Fix issues causing regressions on some targets.
A mismatch between how the proto considers the worst-case-throughput flag to be unset and the way the SchedulingOptions struct does led to the initial version of this incorrectly believing the WCT was unspecified in some scenarios when it was fully set.
Original Description: Refactor XLS schedulers to use a common interface and allow for selecting scheduler for minimization workloads.
This change introduces a
Schedulerabstract base class, and refactorsASAP,MinCut, andSDCscheduling algorithms to implement this interface. A newRandomScheduleris also added. Therun_pipeline_scheduleutility is updated to use this new interface. Additionally, a new option and flag--find_bounds_strategyare added, allowing a different scheduler (defaulting to the selected scheduling algorithm usually SDC) to be used for finding minimum clock period and worst-case throughput bounds, which can be faster than using the main scheduling strategy. Error handling inrun_pipeline_scheduleis also improved.The 'random' scheduler is also updated to make use of the ASAP scheduler bounds to do a much better job at consistently choosing allowable schedules.
The previous run_pipeline_scheduleing had a number of subtle bugs around how its configuration was interpreted that have all been fixed up, requiring some changes to tests which relied on the previous behavior.
Testing is improved to check schedulers other than SDC. Many tests needed to be tweaked to account for the different ways these schedulers work.