demo: self-contained PR benchmark action#88

Merged

afermg merged 23 commits into

mainfrom

demo/benchmark

Jun 17, 2026

Commits on Jun 17, 2026

feat(synth): deterministic synthetic cell-image generator for benchmarking

authored and
afermg
committed
refactor(synth): review hardening — single cell-extent, sturdier tests, determinism

authored and
afermg
committed
feat(bench): symbol-level PR target mapper (build step 2)

authored and
afermg
committed
revert(bench): drop the change-detection mapper; benchmark all exposed functions

authored and
afermg
committed
feat(bench): fixture/runner/comparator — benchmark all get_* head-vs-main

authored and
afermg
committed
feat(bench): two-job benchmark workflow + orchestration driver (build step 3)

authored and
afermg
committed
refactor(bench): review cleanup — leaner comments/docstrings + small fixes

authored and
afermg
committed
test(bench): trim to a lean regression set

timtreis
authored and
afermg
committed
style: ruff-format with current ruff (88-col line wraps)

timtreis
authored and
afermg
committed
chore: stop tracking scratch tasks/ and .claude/ (added by mistake)

timtreis
authored and
afermg
committed
refactor(bench): report raw main/head timings, drop noise-band classification

timtreis
authored and
afermg
committed
Revert "Merge pull request #80 from afermg/feat/synth-bench-generator"

timtreis
authored and
afermg
committed
feat(bench): synthetic-image PR performance benchmark action

timtreis
authored and
afermg
committed
Revert "feat(bench): synthetic-image PR performance benchmark action"

timtreis
authored and
afermg
committed
demo: self-contained PR benchmark action (simplified)

timtreis
authored and
afermg
committed
demo: move the whole benchmark into .github/scripts (no package module)

timtreis
authored and
afermg
committed
demo: extend benchmark matrix to 4 sizes x 2 counts (256–2048)

timtreis
authored and
afermg
committed
demo: median per cell, 3 seeds x 3 counts, dynamic affected-threshold caption

timtreis
authored and
afermg
committed
demo: drop 256px image size (unrealistically small)

timtreis
authored and
afermg
committed
demo: shift matrix down to 256-1024 (drop slow 2048)

timtreis
authored and
afermg
committed
demo: report regressions too — flag functions that moved >=1.05x either way

timtreis
authored and
afermg
committed
demo: slim benchmark.py — drop unused bits

timtreis
authored and
afermg
committed
revert(granunlarity): it has an independent PR, was used as test
afermg
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

demo: self-contained PR benchmark action#88

demo: self-contained PR benchmark action#88
afermg merged 23 commits into
mainfrom
demo/benchmark

Commits on Jun 17, 2026

feat(synth): deterministic synthetic cell-image generator for benchmarking

refactor(synth): review hardening — single cell-extent, sturdier tests, determinism

feat(bench): symbol-level PR target mapper (build step 2)

revert(bench): drop the change-detection mapper; benchmark all exposed functions

feat(bench): fixture/runner/comparator — benchmark all get_* head-vs-main

feat(bench): two-job benchmark workflow + orchestration driver (build step 3)

refactor(bench): review cleanup — leaner comments/docstrings + small fixes

test(bench): trim to a lean regression set

style: ruff-format with current ruff (88-col line wraps)

chore: stop tracking scratch tasks/ and .claude/ (added by mistake)

refactor(bench): report raw main/head timings, drop noise-band classification

Revert "Merge pull request #80 from afermg/feat/synth-bench-generator"

feat(bench): synthetic-image PR performance benchmark action

Revert "feat(bench): synthetic-image PR performance benchmark action"

demo: self-contained PR benchmark action (simplified)

demo: move the whole benchmark into .github/scripts (no package module)

demo: extend benchmark matrix to 4 sizes x 2 counts (256–2048)

demo: median per cell, 3 seeds x 3 counts, dynamic affected-threshold caption

demo: drop 256px image size (unrealistically small)

demo: shift matrix down to 256-1024 (drop slow 2048)

demo: report regressions too — flag functions that moved >=1.05x either way

demo: slim benchmark.py — drop unused bits

revert(granunlarity): it has an independent PR, was used as test