Conversation
* allow input slots > output slots in slot transfer * implementing decrypt function in ring_gsw (not perfect yet) * implement decrypt circuit for ring gsw
…_goldreich_ring_gsw_bench
…imize_preimage_memory
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: a409b532c8
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| struct BenchSummaryCacheKey { | ||
| circuit_ptr: usize, | ||
| } |
There was a problem hiding this comment.
The cache key now uses only circuit_ptr, so two calls to the same sub-circuit with different param_bindings share one cached CircuitBenchSummary. That is incorrect because summaries are computed via estimate_gate_bench_with_bindings, which resolves gate parameters from those bindings; when binding values change estimator behavior, the second call silently reuses the first call’s time/latency/parallelism and under/overestimates the circuit.
Useful? React with 👍 / 👎.
| cached.modified == modified && | ||
| cached.len == len |
There was a problem hiding this comment.
This cache hit condition trusts only (modified,len), so lookup_tables.index rewrites with identical file size inside one filesystem timestamp tick will be treated as unchanged and return stale entries. In that case reads can resolve to outdated offsets/keys and load the wrong matrix bytes until metadata changes again, which can corrupt lookup-table reads in long-running processes.
Useful? React with 👍 / 👎.
1 participant
No description provided.