Help Test MiniRacer.md

MiniRacer https://github.com/discourse/mini_racer is ready for testing.

Can you help out measuring how it improves performance for rake assets:precompile in your app?

• Run:

rm -fr public/assets
RAILS_ENV=production time bundle exec rake assets:precompile

• Edit Gemfile. Remove therubyracer, add:

gem 'execjs', github: 'rails/execjs'
gem 'mini_racer'

• bundle install

• Run:

rm -fr public/assets
RAILS_ENV=production time bundle exec rake assets:precompile

I also did a rm -rf tmp/assets before each run.

133s before, 105s after.

System Node took 40 seconds vs 30 seconds using MiniRacer

Edit: this is the command I used:

rm -rf tmp public/assets && time rake assets:precompile

I think there is a way to run compilation in parallel, people with long running times should try that as well as MiniRacer is safe to use from multiple threads (you get an isolate per context)

In our largest app:

system nodejs
real  1m40.834s      user  1m29.940s      sys 0m11.895s

therubyracer
real 1m14.872s       user 1m13.072s       sys 0m2.153s

mini_racer
real 0m50.688s       user 0m51.122s       sys 0m2.688s

A smaller app:

system nodejs
real 0m28.631s       user 0m25.277s       sys 0m3.931s

therubyracer
real 0m6.153s        user 0m5.762s        sys 0m0.513s

mini_racer
real 0m6.659s        user 0m6.951s        sys 0m0.594s

Big old 3.2 app, with therubyracer:

54.21 real        48.66 user         4.48 sys

With mini_racer:

36.70 real        29.15 user         5.14 sys

System nodejs

57.44s user 5.90s system 101% cpu 1:02.34 total

mini_racer

7.08s user 3.48s system 89% cpu 11.824 total

System nodejs

2.65s user 0.84s system 101% cpu 3.436 total

mini_racer

2.61s user 0.82s system 101% cpu 3.380 total

therubyracer:

61.20 real        54.50 user         2.17 sys

mini_racer: 6x faster! :-)

10.30 real         5.26 user         1.50 sys

ruby 2.0.0p481 on OSX 10.11.4

nodejs
9.37s user 1.59s system 99% cpu 11.065 total
9.58s user 1.62s system 99% cpu 11.310 total
9.49s user 1.60s system 99% cpu 11.199 total

mini_racer
9.21s user 1.59s system 99% cpu 10.894 total
9.26s user 1.60s system 99% cpu 10.920 total
9.07s user 1.57s system 99% cpu 10.691 total

Before (therubyracer)

265.78s user 14.32s system 96% cpu 4:50.77 total

After (mini_racer)

263.41s user 13.61s system 97% cpu 4:43.23 total

Rails 4.1.15

using nodejs 4.0.0
76.85s user 19.09s system 86% cpu 1:51.19 total

using mini_racer
14.65s user 6.44s system 68% cpu 30.843 total

🎉

Are you guys deleting the sprockets cache (tmp/assets) between runs? I got several times speed up at first too but then after doing that the speed up was reduced to 25% (still good, but not as unbelievable as I first thought).

Before (therubracer)
21.35s user 6.19s system 96% cpu 28.639 total

After (mini_racer)
4.67s user 1.82s system 84% cpu 7.663 total

Running in a small Rails 5 application (ruby 2.3.1p112) the command:
time RAILS_ENV=production bundle exec rake assets:precompile
(after running rm -rf public/assets tmp/cache/assets)

therubyracer
30.90s user 0.91s system 100% cpu 31.744 total

system node (v4.4.4)
13.35s user 0.99s system 103% cpu 13.913 total

mini_racer
12.33s user 1.02s system 103% cpu 12.858 total

Are you guys deleting the sprockets cache (tmp/assets) between runs? I got several times speed up at first too but then after doing that the speed up was reduced to 25% (still good, but not as unbelievable as I first thought).

I've been using assets:clobber

Results after running rm -rf public/assets tmp/cache/assets before runs:

therubyracer:
67.42s user 3.39s system 95% cpu 1:14.36 total

mini_racer:
55.45s user 3.28s system 97% cpu 59.960 total

Cleared caches and made sure spring wasn't running. Results:

therubyracer: 3m50s user
mini_racer: 1m48s user