Skip to content

fix(benchmarks): keep several round fold labels on the sweep colourbar at small --clip#32

Merged
FBumann merged 1 commit into
masterfrom
fix/sweep-fold-ticks
Jun 6, 2026
Merged

fix(benchmarks): keep several round fold labels on the sweep colourbar at small --clip#32
FBumann merged 1 commit into
masterfrom
fix/sweep-fold-ticks

Conversation

@FBumann

@FBumann FBumann commented Jun 6, 2026

Copy link
Copy Markdown
Collaborator

TODO (human): one line on why.

Note

Generated by AI (Claude).

What

The sweep heatmap's fold colourbar placed ticks at integer log₂ steps (…, 1/2×, , , …). For a tight colour range — a small --clip like 1.5 (bound = log₂1.5 ≈ 0.585), or a naturally tight p95 default — those integer steps fall outside [-bound, bound], so only rendered.

Now ticks fit the range and read as round folds: clean powers of two for a wide range, round sub-2× folds (1.5×, 1.25×, 1.1×) for a tight one, with an even-spaced fallback when nothing round fits.

--clip 1.51/1.5×, 1/1.25×, 1/1.1×, 1×, 1.1×, 1.25×, 1.5× instead of a lone .

Verify

  • ruff + mypy clean; _fold_ticks gives in-range, symmetric labels across --clip 1.1 / 1.5 / 2 / 8 and default p95 bounds.

…mall --clip

The fold colourbar placed ticks at integer log2 steps (…, 1/2×, 1×, 2×, …). For
a tight range — a small `--clip` like 1.5 (bound = log2 1.5 ≈ 0.585), or a
naturally tight p95 default — those integer steps fall *outside* [-bound, bound],
so only `1×` rendered.

Pick ticks that fit the range and read as round folds: clean powers of two for a
wide range, round sub-2× folds (1.5×, 1.25×, 1.1×) for a tight one, with an
even-spaced fallback when nothing round fits. `--clip 1.5` now shows
1/1.5× … 1.5× instead of a lone 1×.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@FBumann FBumann merged commit 4db3c76 into master Jun 6, 2026
15 of 17 checks passed
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.

1 participant