Repository navigation
Compacted tracing: step budgets that drop finished particles - #236
Draft
rogeriojorge wants to merge 14 commits into
Draft
rogeriojorge wants to merge 14 commits into
rogeriojorge wants to merge 14 commits into
Conversation
…ts resume from their last save
…on the device once
…d-step controller state)
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Limit the steps spent on each round of batched CPU tracing. The additional bookkeeping was slower on GPU, so the device path ignores the step budget.
On the office CPU a VMEC right-hand-side call took 0.74 ms for eight particles and 1.6 ms for 64. The benefit depends on how many lanes finish early and how the field evaluation scales with batch size.
Measurements and test conditions