Skip to content

Speedup package:crypto (with a focus on md5) #892

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 9 commits into
base: main
Choose a base branch
from

Conversation

jensjoha
Copy link
Contributor

TL;DR
This makes it faster and use less ram.

Details
Tested on out/ReleaseX64/wasm-opt from a dart sdk build (167MB):

JIT, dart --deterministic:

Difference at 95.0% confidence
        -313.4 +/- 44.2824
        -22.8359% +/- 3.22664%
        (Student's t, pooled s = 30.3628)

JIT, dart:

Difference at 95.0% confidence
        -275.6 +/- 43.5498
        -20.1905% +/- 3.19046%
        (Student's t, pooled s = 29.8605)

AOT, dartaotruntime --deterministic:

Difference at 95.0% confidence
        -499 +/- 11.5944
        -40.3135% +/- 0.936693%
        (Student's t, pooled s = 7.94984)

AOT, dartaotruntime:

Difference at 95.0% confidence
        -506.8 +/- 39.0824
        -41.304% +/- 3.1852%
        (Student's t, pooled s = 26.7974)

Running the AOT snapshots through the "benchmarker" utility from dart front_end I get these results (with 25 runs of each version):

msec task-clock:u: -41.1276% +/- 0.3390% (-515.19 +/- 4.25) (1252.67 -> 737.48)
page-faults:u: -55.7760% +/- 0.1266% (-655.48 +/- 1.49) (1175.20 -> 519.72)
cycles:u: -41.7236% +/- 0.3140% (-2185317986.60 +/- 16443529.36) (5237610655.96 -> 3052292669.36)
instructions:u: -32.5192% +/- 0.0000% (-5346525546.76 +/- 2905.52) (16441121298.68 -> 11094595751.92)
branch-misses:u: -97.0502% +/- 0.3379% (-8025400.44 +/- 27942.52) (8269332.32 -> 243931.88)
seconds time elapsed: -41.0827% +/- 0.3407% (-0.52 +/- 0.00) (1.25 -> 0.74)
seconds user: -41.7571% +/- 0.4373% (-0.50 +/- 0.01) (1.19 -> 0.70)
seconds sys: -27.8724% +/- 4.1716% (-0.02 +/- 0.00) (0.06 -> 0.04)
MarkSweep(   old space) goes from 2 to 1

Running it through /usr/bin/time -v reports the "Maximum resident set size (kbytes)" going from around ~346k to ~176k (i.e. uses about half the ram).

MD5 calculations is used internally in the analyzer, and patching this in, for an AOT compile of the analyzer we see changes like this (when run through the benchmarker with parameters like --train-using pkg/front_end/lib/ in the dart sdk):

msec task-clock:u: -1.2865% +/- 0.2432% (-105.99 +/- 20.03) (8238.61 -> 8132.62)
page-faults:u: -6.7334% +/- 0.0969% (-10705.66 +/- 154.06) (158992.30 -> 148286.64)
cycles:u: -1.1201% +/- 0.2487% (-384785056.06 +/- 85428229.56) (34351469228.82 -> 33966684172.76)
instructions:u: -2.6679% +/- 0.0191% (-1227909335.74 +/- 8796243.14) (46025301866.14 -> 44797392530.40)
branch-misses:u: -1.5364% +/- 0.9908% (-1886562.18 +/- 1216638.62) (122788764.20 -> 120902202.02)
seconds time elapsed: -1.3342% +/- 0.2478% (-0.11 +/- 0.02) (8.27 -> 8.16)
seconds user: -1.0854% +/- 0.2985% (-0.09 +/- 0.02) (7.88 -> 7.79)
seconds sys: -5.8743% +/- 3.7392% (-0.02 +/- 0.01) (0.36 -> 0.34)
Scavenge(   new space) goes from 134 to 131
MarkSweep(   old space) goes from 3 to 2

jensjoha added 6 commits May 19, 2025 12:29
An aot-compile, hashing a big file says

```
Difference at 95.0% confidence
	-140.4 +/- 35.1436
	-11.2825% +/- 2.82414%
	(Student's t, pooled s = 24.0967)
```
@jensjoha jensjoha requested a review from a team as a code owner May 22, 2025 13:41
Copy link

github-actions bot commented May 22, 2025

PR Health

Breaking changes ✔️
Package Change Current Version New Version Needed Version Looking good?
crypto None 3.0.6 3.0.7-wip 3.0.6 ✔️
Changelog Entry
Package Changed Files
package:crypto pkgs/crypto/lib/src/hash_sink.dart
pkgs/crypto/lib/src/md5.dart

Changes to files need to be accounted for in their respective changelogs.

This check can be disabled by tagging the PR with skip-changelog-check.

Coverage ⚠️
File Coverage
pkgs/crypto/benchmark/benchmark.dart 💔 Not covered
pkgs/crypto/lib/src/hash_sink.dart 💚 98 % ⬆️ 0 %
pkgs/crypto/lib/src/md5.dart 💚 100 %
pkgs/crypto/tool/md5sum.dart 💔 Not covered

This check for test coverage is informational (issues shown here will not fail the PR).

This check can be disabled by tagging the PR with skip-coverage-check.

API leaks ✔️

The following packages contain symbols visible in the public API, but not exported by the library. Export these symbols or remove them from your publicly visible API.

Package Leaked API symbols
License Headers ✔️
// Copyright (c) 2025, the Dart project authors. Please see the AUTHORS file
// for details. All rights reserved. Use of this source code is governed by a
// BSD-style license that can be found in the LICENSE file.
Files
no missing headers

All source files should start with a license header.

@mosuem mosuem requested a review from lrhn May 22, 2025 14:18
Copy link
Member

@lrhn lrhn left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM.
Feel free to ignore my style comments if it works and is faster, and you can't be bothered.
We can always improve more later.

} else if (arg == 'sha512256') {
setFunction(sha512256, 'sha512/256');
} else {
throw ArgumentError('Unknown argument: $arg');
Copy link
Member

@lrhn lrhn May 23, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe print an error on stderr and exit. This is a command line usage mistake, not a Dart argument error.

print('Using hash function $message');
}

for (var arg in args) {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Seems like you only need to look at args.first. If there are two, either one is an error, or the second will find the function already set.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants