Skip to content

Measure and validate scalar replay mux allocation savings - #542

Merged
xtofalex merged 1 commit into
mainfrom
measure/pr541
Nov 17, 2026
Merged

xtofalex merged 1 commit into
mainfrom
measure/pr541

Conversation

@xtofalex

Copy link
Copy Markdown
Collaborator

PR #541 had only synthetic structural evidence. This report compares its exact parent (49bfc050) with the merged revision (1689438c, freshly fetched main for the campaign), using matching Release builds and the established 13-case public/synthetic corpus. Production code is unchanged and both frontend cleanup calls remain enabled.

Direct diagnostic counters show that #541 avoids all 207 targeted scalar replay mux allocations / 207 output bits on Ibex. Cleanup still directly removes 176 muxes / 744 bits from Ibex and 18 / 340 from AES VHDL. The report separates cumulative allocations, live cleanup counts, folded and hierarchy-expanded structure, and independently measured total instances.

The separate uninstrumented campaign retains all 312 balanced paired samples after 26 declared warm-ups. Ibex's median paired loading-time change is -1.64% (exploratory 95% interval -3.44% to -0.86%); no representative RSS improvement is established. Exact environment, revisions, methods, limitations and sanitized CSV evidence are included.

Validation:

  • CMake frontend/intent, VHDL, HDL helper, mux-utility, serialization and cross-process determinism checks passed on both revisions.
  • Source-based Bazel dependency selection passed, but compilation failed the local Xcode SDK absolute-include check for MacOSX.sdk/SDKSettings.json; zero Bazel tests executed. Build-path parity remains unvalidated.
  • REUSE lint and git diff --check passed.

Recommendation: keep both cleanup calls. Separately assess eligibility of the 26 remaining five-bit same-case replay pairs before considering another optimization. No vector optimization, broad planner, dead-RHS work or primitive changes are included.

@codspeed

codspeed Bot commented Nov 17, 2026

Copy link
Copy Markdown

Merging this PR will not alter performance

1 warning

⚠️ 13 benchmarks are dominated by system calls

System calls cannot be simulated consistently and are left out of the measure. The reported value understates the real cost. Switch to the Walltime instrument to measure them.

Learn more about system calls

See affected benchmarks

✅ 111 untouched benchmarks


Comparing measure/pr541 (b07f1bd) with main (1689438)

Open in CodSpeed

@codecov-commenter

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

@xtofalex
xtofalex merged commit c9563f3 into main Nov 17, 2026
65 of 66 checks passed
@xtofalex
xtofalex deleted the measure/pr541 branch November 17, 2026 16:16
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