search

zenika-open-source / immutadot

immutadot is a JavaScript library to deal with nested immutable structures.
https://immutadot.zenika.com
MIT License
178 stars 5 forks source link

Add more benchmark tests and seamless immutable #299

Open nlepage opened 6 years ago

nlepage commented 6 years ago

Some ideas:

nlepage commented 6 years ago

New output format in order to display more results:

ES2015 destructuring immutable 3.8.2 immer 1.2.0 qim 0.0.52 immutad●t 2.0.0
Update small todos list (1000 items) 14841ops/s (0.07ms/op) 5961ops/s (0.17ms/op) 2668ops/s (0.37ms/op) 10515ops/s (0.10ms/op) 11837ops/s (0.08ms/op)
Update medium todos list (10000 items) 1567ops/s (0.64ms/op) 554ops/s (1.81ms/op) 270ops/s (3.70ms/op) 1095ops/s (0.91ms/op) 1361ops/s (0.73ms/op)
Update large todos list (100000 items) 119ops/s (8.40ms/op) 54ops/s (18.54ms/op) 24ops/s (40.89ms/op) 91ops/s (11.02ms/op) 106ops/s (9.44ms/op)
nlepage commented 6 years ago

Also allow to run only one benchmark with yarn start setProp or yarn fast setProp. This is not so easy with jest since each test file is isolated...

nlepage commented 5 years ago

Also allow to run only one benchmark with yarn start setProp or yarn fast setProp. This is not so easy with jest since each test file is isolated...

Tried doing this with jest, got some headaches... so no.

nlepage commented 5 years ago

Latest output:

ES2015 destructuring immutable 3.8.2 seamless-immutable 7.1.3 immer 1.2.0 qim 0.0.52 immutad●t 2.0.0 qim 0.0.52 curried immutad●t 2.0.0 curried
Set a property 5868546ops/s (170ns/op) 518726ops/s (1.93µs/op) 67781ops/s (14.8µs/op) 165929ops/s (6.03µs/op) 883236ops/s (1.13µs/op) 980969ops/s (1.02µs/op) 829188ops/s (1.21µs/op) 907277ops/s (1.10µs/op)
Update small todos list (1000 items) 75077ops/s (13.3µs/op) 5769ops/s (173µs/op) 580ops/s (1.72ms/op) 2610ops/s (383µs/op) 17674ops/s (56.6µs/op) 24510ops/s (40.8µs/op) 17941ops/s (55.7µs/op) 26289ops/s (38.0µs/op)
Update medium todos list (10000 items) 7185ops/s (139µs/op) 555ops/s (1.80ms/op) 60ops/s (16.8ms/op) 259ops/s (3.86ms/op) 1886ops/s (530µs/op) 3354ops/s (298µs/op) 1884ops/s (531µs/op) 3354ops/s (298µs/op)
Update large todos list (100000 items) 286ops/s (3.50ms/op) 48ops/s (21.0ms/op) 6ops/s (181ms/op) 24ops/s (41.6ms/op) 135ops/s (7.38ms/op) 193ops/s (5.19ms/op) 130ops/s (7.69ms/op) 204ops/s (4.90ms/op)
Final score 100 10 1 5 28 41 27 42

Performance improvements were made on apply and curry, so this is even more important, in order to validate those improvements.

Also, could be nice to have the score on each line.

nlepage commented 5 years ago

https://nodesource.com/blog/tracking-down-performance-bottlenecks-nsolid-deoptigate

nlepage commented 5 years ago

Soooooo... Right now on master (b6b470e) I get this :

ES2015 destructuring immutable 3.8.2 seamless-immutable 7.1.4 immer 1.12.1 qim 0.0.52 immutad●t 2.0.0 qim 0.0.52 curried immutad●t 2.0.0 curried
Set a property 5555557ops/s (180ns/op) 171347ops/s (5.84µs/op) 59411ops/s (16.8µs/op) 108341ops/s (9.23µs/op) 880437ops/s (1.14µs/op) 808669ops/s (1.24µs/op) 827541ops/s (1.21µs/op) 748391ops/s (1.34µs/op)
Set a deeply nested property 961724ops/s (1.04µs/op) 22309ops/s (44.8µs/op) 4789ops/s (209µs/op) 10879ops/s (91.9µs/op) 78197ops/s (12.8µs/op) 53322ops/s (18.8µs/op) 69412ops/s (14.4µs/op) 57379ops/s (17.4µs/op)
Update small todos list (1000 items) 6611ops/s (151µs/op) 1465ops/s (682µs/op) 416ops/s (2.40ms/op) 1242ops/s (805µs/op) 4356ops/s (230µs/op) 19795ops/s (50.5µs/op) 3931ops/s (254µs/op) 18376ops/s (54.4µs/op)
Update medium todos list (10000 items) 599ops/s (1.67ms/op) 125ops/s (7.99ms/op) 44ops/s (23.0ms/op) 120ops/s (8.34ms/op) 439ops/s (2.28ms/op) 3064ops/s (326µs/op) 488ops/s (2.05ms/op) 3004ops/s (333µs/op)
Update large todos list (100000 items) 58ops/s (17.4ms/op) 14ops/s (69.7ms/op) 5ops/s (209ms/op) 13ops/s (79.2ms/op) 44ops/s (22.6ms/op) 196ops/s (5.10ms/op) 44ops/s (22.7ms/op) 195ops/s (5.13ms/op)
Final score 100 15 5 13 48 234 48 228

On Update todos, it might be a false positive... I'll investigate.

nlepage commented 5 years ago

So going back before rewriting the parser in TS (551e60d) it goes back to former results... This would mean that somehow, rewriting the parser in TS gave a huge performance gain on slices! Makes no sense... @hgwood

nlepage commented 5 years ago

And it's faster than ES2015 :upside_down_face:

nlepage commented 5 years ago

See #332 ES2015 updateTodos was slow...