[lesnik512]

Regenerates the comparison tables in docs/introduction/performance.md from main at 473ce2a, so they include #694 (each factory gets its own copy of its resolver code).

C3 (six-deep chain) dropped from about 760 ns to 552-557 ns. Against dishka, modern-di's C3 ratio went from 1.21 to 0.87, so modern-di is now faster on every scenario except the warm singleton lookup. The other cells moved by a few percent at most.

Changes:

• The tables are pasted unedited from just bench-report (5 runs).
• The intro now says modern-di is slower only on the warm singleton lookup. The request-cycle figure is 1.66 µs instead of 1.74 µs, here and in comparison.md.
• The measurement line has the new date, commit and baseline (4.1.0 plus unreleased main). The machine, Python and rival versions are unchanged.
• The C2/C3 caveat drops "or why dishka wins C3". Two numbers in the caveats follow the new data: dependency-injector's C2 is 67 ns, and the C4 event-loop floor is about 18% of modern-di's time.

The machine had a load average of about 2.8 and was about 85% idle during the run. modern-di spreads are within 1.7% (previously 1.4%) and rival spreads within 4.7% (previously 3.8%). The widest cell is C6 against wireup at ±4.2%. All C3 cells are within 1.5%. Earlier runs on a busier machine had wider spreads, but their medians matched these to within 1-2%, and the C3 ratio against dishka was 0.87-0.88 every time.

Checked: just lint-ci, just test-docs (218 passed), tests/test_bench_report.py, and just docs-build.

[github-actions]

Benchmark

Details

Benchmark suite	Current: 54f1e1c	Previous: 1fe9404	Ratio
benchmarks/test_guard_by_type.py::test_g16_resolve_by_type	3320957.012132102 iter/sec (stddev: 1.1938086467269671e-8)	5178696.623565742 iter/sec (stddev: 5.414322976795206e-9)	1.56
benchmarks/test_guard_by_type.py::test_g17_resolve_by_type_large_registry	3331223.78035057 iter/sec (stddev: 1.4790890137353807e-8)	5173048.561866323 iter/sec (stddev: 4.944868870356742e-9)	1.55
benchmarks/test_guard_cold.py::test_g8_cold_first_resolve	16023.527758702754 iter/sec (stddev: 0.000052008717808257516)	23341.89100892437 iter/sec (stddev: 0.000040957484774952956)	1.46
benchmarks/test_guard_cold.py::test_g8b_cold_first_resolve_cached	12713.762094084948 iter/sec (stddev: 0.0001280532705323862)	18128.971481425186 iter/sec (stddev: 0.000230226555941953)	1.43
benchmarks/test_guard_concurrency.py::test_g14_concurrent_cached_hit[1]	472.5629558719837 iter/sec (stddev: 0.000050515451875524175)	500.63779365711235 iter/sec (stddev: 0.00009568522495976157)	1.06
benchmarks/test_guard_concurrency.py::test_g14_concurrent_cached_hit[2]	469.9766829201808 iter/sec (stddev: 0.000055153807276442504)	557.5535854996675 iter/sec (stddev: 0.00010071569364556589)	1.19
benchmarks/test_guard_concurrency.py::test_g14_concurrent_cached_hit[4]	403.82844280480515 iter/sec (stddev: 0.00013292801451865268)	569.4122020523743 iter/sec (stddev: 0.00016102760266743736)	1.41
benchmarks/test_guard_concurrency.py::test_g15_concurrent_first_resolve[1]	13476.606127693349 iter/sec (stddev: 0.00000759092088704966)	18760.14220174014 iter/sec (stddev: 0.000007857094267724099)	1.39
benchmarks/test_guard_concurrency.py::test_g15_concurrent_first_resolve[2]	6345.95379609425 iter/sec (stddev: 0.000020953630848694298)	9555.709667947016 iter/sec (stddev: 0.000018284080800311374)	1.51
benchmarks/test_guard_concurrency.py::test_g15_concurrent_first_resolve[4]	3329.988267618312 iter/sec (stddev: 0.00004192420271874652)	4955.91200034446 iter/sec (stddev: 0.00003013749565134662)	1.49
benchmarks/test_guard_concurrency.py::test_g15b_concurrent_first_resolve_sibling_children[1]	9976.021800224114 iter/sec (stddev: 0.000009628038301447758)	14076.259543718805 iter/sec (stddev: 0.00000931467062375488)	1.41
benchmarks/test_guard_concurrency.py::test_g15b_concurrent_first_resolve_sibling_children[2]	4183.459333861973 iter/sec (stddev: 0.000017172115705238353)	5964.930088272373 iter/sec (stddev: 0.00002430283539827365)	1.43
benchmarks/test_guard_concurrency.py::test_g15b_concurrent_first_resolve_sibling_children[4]	2010.368879564 iter/sec (stddev: 0.00005464951550704635)	2879.198822617753 iter/sec (stddev: 0.00011444513604976574)	1.43
benchmarks/test_guard_concurrency.py::test_g15c_worker_pool_floor_control[1]	39496.82363229264 iter/sec (stddev: 0.0000048353516943913426)	53280.05327926116 iter/sec (stddev: 0.0000025629815202707525)	1.35
benchmarks/test_guard_concurrency.py::test_g15c_worker_pool_floor_control[2]	13882.954889705179 iter/sec (stddev: 0.000013290440313554477)	21659.336184542513 iter/sec (stddev: 0.000008141698894717914)	1.56
benchmarks/test_guard_concurrency.py::test_g15c_worker_pool_floor_control[4]	5681.8711803694405 iter/sec (stddev: 0.00003530451251352048)	8611.282200507329 iter/sec (stddev: 0.000020640219845949377)	1.52
benchmarks/test_guard_lifecycle.py::test_g6_build_child_container	1580052.2823424013 iter/sec (stddev: 3.175697902027839e-8)	2206752.6410179175 iter/sec (stddev: 1.7192874583285688e-8)	1.40
benchmarks/test_guard_lifecycle.py::test_g6b_build_child_container_auto_scope	1687703.3260639743 iter/sec (stddev: 3.165939020217187e-8)	2363877.0278010173 iter/sec (stddev: 1.894917431566577e-8)	1.40
benchmarks/test_guard_lifecycle.py::test_g7_request_lifecycle_batch	3112.708671961348 iter/sec (stddev: 0.00001596420956953311)	4439.610302361827 iter/sec (stddev: 0.00000816229810846762)	1.43
benchmarks/test_guard_lifecycle.py::test_g7c_event_loop_floor_control	57724.27277025886 iter/sec (stddev: 0.0000018466475476650607)	75806.40753304795 iter/sec (stddev: 0.0000013101134732601992)	1.31
benchmarks/test_guard_lifecycle.py::test_g7b_request_cycle_sync	411245.5676206241 iter/sec (stddev: 8.85694235703082e-8)	581927.8974458397 iter/sec (stddev: 6.294860146637674e-8)	1.42
benchmarks/test_guard_lifecycle.py::test_g13_teardown_at_scale	65077.86985603879 iter/sec (stddev: 0.0000015488741927788382)	94302.42871685681 iter/sec (stddev: 0.0000010347080417754609)	1.45
benchmarks/test_guard_lifecycle.py::test_g13b_teardown_at_scale_async_no_finalizers	598.6968247486334 iter/sec (stddev: 0.000894760853536026)	826.6166310205344 iter/sec (stddev: 0.0011973141997795621)	1.38
benchmarks/test_guard_resolve.py::test_g1_transient_resolve	2160796.4220575644 iter/sec (stddev: 3.726822866003395e-8)	3000560.8798204404 iter/sec (stddev: 2.072957044357824e-8)	1.39
benchmarks/test_guard_resolve.py::test_g2_cached_resolve	3614360.06940675 iter/sec (stddev: 9.973134000850693e-9)	5069344.578366031 iter/sec (stddev: 6.151691670543562e-9)	1.40
benchmarks/test_guard_resolve.py::test_g3_deep_chain	782061.8321164199 iter/sec (stddev: 4.228180980032576e-8)	1098187.3073185228 iter/sec (stddev: 4.228505498554438e-8)	1.40
benchmarks/test_guard_resolve.py::test_g4_wide_resolve	466410.8511975815 iter/sec (stddev: 3.9888615274337577e-7)	609703.083134297 iter/sec (stddev: 3.6902082541952153e-7)	1.31
benchmarks/test_guard_resolve.py::test_g5_cross_scope	1991335.9957706158 iter/sec (stddev: 2.3053799842117716e-8)	2606734.7076500254 iter/sec (stddev: 2.6237384655609e-8)	1.31
benchmarks/test_guard_resolve.py::test_g9_context_resolve	1410167.9809713834 iter/sec (stddev: 1.5103813896160022e-7)	1994396.9412285434 iter/sec (stddev: 1.473883486223868e-7)	1.41
benchmarks/test_guard_resolve.py::test_g12_override_active_resolve	829698.8719836429 iter/sec (stddev: 4.9745572981974815e-8)	1159115.7615864046 iter/sec (stddev: 4.834304288025477e-8)	1.40
benchmarks/test_guard_resolve.py::test_g18_alias_hop	3610433.763633581 iter/sec (stddev: 1.3868370165643838e-8)	5004418.150554741 iter/sec (stddev: 4.873760874431247e-9)	1.39
benchmarks/test_guard_validate.py::test_g10_validate_deep_chain	35403.2120202192 iter/sec (stddev: 0.000018852325711566874)	53893.052686058494 iter/sec (stddev: 0.000015060637075350605)	1.52
benchmarks/test_guard_validate.py::test_g11_validate_wide	21224.238258441233 iter/sec (stddev: 0.000023332802985458138)	31556.14456620828 iter/sec (stddev: 0.000019401454019783594)	1.49

This comment was automatically generated by workflow using github-action-benchmark.