Restrict unrolling to loops that are nested within less than 4 loops #1595
Conversation
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
as unrolling and analyzing empty loops is cheaper than visiting loops for collecting the nr of the instructions
karoliineh
added
sv-comp
sim642
approved these changes
Oct 11, 2024
For reference, in SV-COMP 2024 we also had a bunch of fixpoint errors: https://sv-comp.sosy-lab.org/2024/results/results-verified/goblint.results.SV-COMP24.table.html#/table?filter=0(0*status*(status(in(ERROR%20%28verify%29)))). |
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Outdated suggestions cannot be applied.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.
This PR addresses the excessive resource usage introduced in bb6f9aa.
As the amount of unrolling grows exponentially with the number of nested loops, we determined what would be a reasonable amount of nestings with a 5s run on the sv-benchmarks and found that unrolling less than 4 loops is a reasonable metric.
On the downside, this PR introduces an issue where some termination tasks do not reach a fixpoint. Unfortunately, I was not able to trace it to the cause, but we made a quick run, and it seems to not impact our overall scores (compared to SV-COMP 2023, we lost 2 termination tasks, but we also currently run out of resources for these).
So, (hopefully) this PR fixes more things than it breaks. Edit: we made a run and, indeed, this PR does not break anything else but fixes the excessive resource usage with 717 tasks going from errors back to unknowns.