-
Notifications
You must be signed in to change notification settings - Fork 534
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
feat: bv_decide inequality regression tests #5714
Conversation
!bench to get a baseline with ac_nf enabled |
Mathlib CI status (docs):
|
Here are the benchmark results for commit 59413c1. Benchmark Metric Change
==================================================
- import Lean task-clock 8.3% (60.2 σ)
- import Lean wall-clock 8.3% (71.8 σ)
+ lake build clean wall-clock -1.4% (-11.0 σ)
- lake config import task-clock 8.2% (11.7 σ)
- lake config import wall-clock 8.2% (10.9 σ)
- unionfind task-clock 6.6% (16.3 σ)
- unionfind wall-clock 6.6% (16.1 σ) |
59413c1
to
16acc72
Compare
!bench and now without ac_nf |
Note that the worst offender has the normalization pass explicitly disabled (because it times out otherwise), so the baseline will be somewhat skewed |
Here are the benchmark results for commit 16acc72.Found no runs to compare against. |
16acc72
to
432d6f8
Compare
This takes a few standalone bitvector problems, about inequalties, from LNSym, and adds them as a benchmark to prevent further regressions with bv_decide.
These problems are particularly interesting, because they've previously had a bad interaction with bv_decides normalization pass, see #5664.