fix(bench): explicit timers don't force high precision measurements #20272
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
bartlomieju
requested changes
Aug 25, 2023
|
@nayeemrmn looks like there's an actual test failure for the test added in this PR. |
|
@nayeemrmn I'm afraid this test is flaky, eg: https://github.com/denoland/deno/actions/runs/5990043261/job/16246927846?pr=18642 from #18642 |
bartlomieju
pushed a commit
that referenced
this pull request
Aug 31, 2023
…20272) Disables `BenchContext::start()` and `BenchContext::end()` for low precision benchmarks (less than 0.01s per iteration). Prints a warning when they are used in such benchmarks, suggesting to remove them. ```ts Deno.bench("noop", { group: "noops" }, () => {}); Deno.bench("noop with start/end", { group: "noops" }, (b) => { b.start(); b.end(); }); ``` Before: ``` cpu: 12th Gen Intel(R) Core(TM) i9-12900K runtime: deno 1.36.2 (x86_64-unknown-linux-gnu) file:///home/nayeem/projects/deno/temp3.ts benchmark time (avg) iter/s (min … max) p75 p99 p995 ----------------------------------------------------------------------------- ----------------------------- noop 2.63 ns/iter 380,674,131.4 (2.45 ns … 27.78 ns) 2.55 ns 4.03 ns 5.33 ns noop with start and end 302.47 ns/iter 3,306,146.0 (200 ns … 151.2 µs) 300 ns 400 ns 400 ns summary noop 115.14x faster than noop with start and end ``` After: ``` cpu: 12th Gen Intel(R) Core(TM) i9-12900K runtime: deno 1.36.1 (x86_64-unknown-linux-gnu) file:///home/nayeem/projects/deno/temp3.ts benchmark time (avg) iter/s (min … max) p75 p99 p995 ----------------------------------------------------------------------------- ----------------------------- noop 3.01 ns/iter 332,565,561.7 (2.73 ns … 29.54 ns) 2.93 ns 5.29 ns 7.45 ns noop with start and end 7.73 ns/iter 129,291,091.5 (6.61 ns … 46.76 ns) 7.87 ns 13.12 ns 15.32 ns Warning start() and end() calls in "noop with start and end" are ignored because it averages less than 0.01s per iteration. Remove them for better results. summary noop 2.57x faster than noop with start and end ```
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Addresses https://discord.com/channels/684898665143206084/688040863313428503/1143302080602325064.
Disables
BenchContext::start()andBenchContext::end()for low precision benchmarks (less than 0.01s per iteration). Prints a warning when they are used in such benchmarks, suggesting to remove them.Before:
After: