Skip to content
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

Make privatization sync Join more precise when threadflag is path-sensitive #1475

Merged
merged 5 commits into from
May 22, 2024

Conversation

sim642
Copy link
Member

@sim642 sim642 commented May 21, 2024

In #1464, just enabling abortUnless caused two of our simple regression tests in sv-benchmarks to time out:

  1. 13-privatized_01-priv_nr_true — This is embarassing...
  2. 13-privatized_25-struct_nr_true

Tracing the first one revealed MUTEX_INITS of relational mutex-meet getting additional unwidened increasing side-effects from the body of __VERIFIER_assert, in particular, the join point there. The sync Join which causes these is unnecessary in SV-COMP where threadflag is path-sensitive, so branched thread creation doesn't need to be accounted for by privatizations.

Therefore, this PR makes such privatizations check that and avoid unnecessary side-effects from sync Join. It fixes the two cases, which are probably somewhat outliers that they got an explosion of contexts with abortUnless.
Really, this could make us both more precise (by side-effecting less crap to MUTEX_INITS etc) and more efficient (not modifying MUTEX_INITS which everything depends on).

@sim642 sim642 added sv-comp SV-COMP (analyses, results), witnesses precision performance Analysis time, memory usage labels May 21, 2024
@sim642 sim642 added this to the SV-COMP 2025 milestone May 21, 2024
@sim642 sim642 requested a review from michael-schwarz May 21, 2024 08:07
@sim642 sim642 marked this pull request as ready for review May 21, 2024 08:12
@sim642 sim642 mentioned this pull request May 21, 2024
3 tasks
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
performance Analysis time, memory usage precision sv-comp SV-COMP (analyses, results), witnesses
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants