Skip to content

kevm-pyk: thread step_timeout through run_prover#2867

Open
Stevengre wants to merge 2 commits into
masterfrom
jh/run-prover-step-timeout
Open

kevm-pyk: thread step_timeout through run_prover#2867
Stevengre wants to merge 2 commits into
masterfrom
jh/run-prover-step-timeout

Conversation

@Stevengre

@Stevengre Stevengre commented Jun 18, 2026

Copy link
Copy Markdown
Contributor

Adds step_timeout to run_prover, forwarded to APRProver (which enforces the per-step budget: interrupt + halve execute_depth + retry, see runtimeverification/k#4930).

step_timeout is only honored on the sequential advance_proof path — parallel_advance_proof ignores it. So when step_timeout is set, run_prover runs sequentially regardless of force_sequential, warning if max_frontier_parallel > 1. Default None is unchanged.

Enables Kontrol's --step-timeout. Making the parallel path honor step_timeout is a pyk follow-up.

@Stevengre Stevengre force-pushed the jh/run-prover-step-timeout branch from 4774b8c to 6a3c8f5 Compare June 18, 2026 14:34
@Stevengre Stevengre requested a review from anvacaru June 18, 2026 15:14
@Stevengre Stevengre self-assigned this Jun 18, 2026
@Stevengre Stevengre marked this pull request as ready for review June 18, 2026 15:15
Comment thread kevm-pyk/src/kevm_pyk/utils.py Outdated
Adds a `step_timeout: int | None = None` parameter to `run_prover` and
forwards it to the `APRProver` constructor, which enforces the per-step
wall-clock budget (interrupt + halve `execute_depth` + retry).

`step_timeout` is only honored on the sequential `advance_proof` path;
`parallel_advance_proof` does not wrap steps with the timeout budget. So
when `step_timeout` is set, `run_prover` runs sequentially regardless of
`force_sequential`, warning if `max_frontier_parallel > 1` that frontier
parallelism is being dropped.

Default `None` keeps the prior behavior (no per-step timeout, parallel
path unchanged).
@Stevengre Stevengre force-pushed the jh/run-prover-step-timeout branch from 6a3c8f5 to 4b7fef4 Compare June 18, 2026 15:49
@Stevengre Stevengre requested a review from ehildenb June 18, 2026 15:51
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants