FastOsc

A Ruby wrapper around rtosc to encode and decode OSC messages.

This also includes a fallback implementation in pure Ruby in the case that the compiled version doesn't load properly. This can be forced by setting an environment variable of FAST_OSC_USE_FALLBACK=1 where needed.

Installation

Add this line to your application's Gemfile:

gem 'fast_osc'

And then execute:

$ bundle

Or install it yourself as:

$ gem install fast_osc

Is it fast?

Let's see...

Key:

fast_osc - this gem
osc - osc-ruby
samsosc - OSC classes from Sonic Pi (which are optimised pure Ruby based on pack and unpack)

Encoding Benchmark

Warming up --------------------------------------
            fast_osc    64.995k i/100ms
             samsosc    23.371k i/100ms
            osc-ruby     7.691k i/100ms
Calculating -------------------------------------
            fast_osc    797.673k (±15.0%) i/s -      3.900M in   5.043770s
             samsosc    258.331k (±12.8%) i/s -      1.285M in   5.063755s
            osc-ruby     83.203k (±12.6%) i/s -    415.314k in   5.073578s

Decoding Bencmark

Warming up --------------------------------------
            fast_osc   102.344k i/100ms
             samsosc    20.770k i/100ms
            osc-ruby     3.145k i/100ms
Calculating -------------------------------------
            fast_osc      1.650M (±14.5%) i/s -      8.085M in   5.017162s
             samsosc    234.951k (±14.0%) i/s -      1.163M in   5.049167s
            osc-ruby     34.266k (±13.3%) i/s -    169.830k in   5.048509s

Benchmarks are now part of this repo - run rake test to see the results for yourself.

What about Truffle Ruby?

A high performance implementation of the Ruby programming language. Built on GraalVM by Oracle Labs.

Just for fun, I re-ran the benchmarks using TruffleRuby. First some install steps:

git clone this repo
bundle install
export PATH="/usr/local/opt/llvm@4/bin:$PATH"
rake clean && rake compile

then the test:

$ rake test
ENCODING TEST
Warming up --------------------------------------
            fast_osc     3.000  i/100ms
             samsosc    28.219k i/100ms
            osc-ruby   234.000  i/100ms
Calculating -------------------------------------
            fast_osc     33.294  (±45.1%) i/s -    129.000  in   5.295649s
             samsosc      1.369M (±33.5%) i/s -      3.556M in   5.010343s
            osc-ruby    274.170k (±20.2%) i/s -      1.064M in   4.971832s
DECODING TEST
Warming up --------------------------------------
            fast_osc     9.000  i/100ms
             samsosc     6.087k i/100ms
            osc-ruby   418.000  i/100ms
Calculating -------------------------------------
            fast_osc     71.034  (±45.0%) i/s -    261.000  in   5.015393s
             samsosc    114.443k (±70.2%) i/s -    261.741k in   5.283892s
            osc-ruby     84.236k (±34.7%) i/s -    237.424k in   5.317738s

# update - retested using rc5
ENCODING TEST
Warming up --------------------------------------
            fast_osc   106.000  i/100ms
             samsosc   999.000  i/100ms
            osc-ruby   159.000  i/100ms
Calculating -------------------------------------
            fast_osc     24.507k (±28.6%) i/s -     83.316k in   4.984544s
             samsosc      1.494M (±22.4%) i/s -      2.785M in   4.988822s
            osc-ruby    265.322k (±26.5%) i/s -    597.522k in   5.014127s
DECODING TEST
Warming up --------------------------------------
            fast_osc   173.000  i/100ms
             samsosc   454.000  i/100ms
            osc-ruby   208.000  i/100ms
Calculating -------------------------------------
            fast_osc     73.712k (±66.4%) i/s -    163.831k in   5.049410s
             samsosc    325.920k (±33.7%) i/s -    875.766k in   4.989279s
            osc-ruby     63.624k (±52.3%) i/s -    129.792k in   5.005088s
Run options: --seed 56032

The Good

The encoding benchmark - the optimised pure Ruby version is nearly as fast as the C extension!

# C ext in MRI
fast_osc    797.673k (±15.0%) i/s -      3.900M in   5.043770s
# Pure Ruby in Truffle rc3
samsosc      1.369M (±33.5%) i/s -      3.556M in   5.010343s
# Pure Ruby in Truffle rc5
samsosc      1.494M (±22.4%) i/s -      2.785M in   4.988822s

The Bad

Decoding was generally slower, although the (non optimised) osc gem seemed to prefer TruffleRuby, running around 40% faster.

~Also the performance of the C extension in TruffleRuby was very, very poor but this may be due to not warming up enough.~ update - this appears to be better in rc5.

Usage

>> FastOsc.encode_single_message("/foo", ["baz", 1, 2.0])
=> "/foo\x00\x00\x00\x00,sif\x00\x00\x00\x00baz\x00\x00\x00\x00\x01@\x00\x00\x00"
>> res = _
>> FastOsc.decode_single_message(res)
=> ["/foo", ["baz", 1, 2.0]]
>> FastOsc.encode_single_bundle(Time.now.to_i, "/foo", ["baz", 1, 2.0])
=> "#bundle\x00\x00\x00\x00\x00W*1\x7F\x00\x00\x00\x1C/foo\x00\x00\x00\x00,sif\x00\x00\x00\x00baz\x00\x00\x00\x00\x01@\x00\x00\x00"

See the test suite for additional methods regarding bundles with timestamps. Bundles are only supported with a single message at present. A timestamp of nil is a special case meaning "immediately".

Running the test suite

$ gem install minitest # or bundle install
$ rake clean && rake clobber && rake compile && rake test

Still todo

[x] Implement more types
[x] Bring benchmarks into the repo
[ ] Work out cross compilation story for easier packaging
[x] Implement multi message/nested bundles
[ ] More documentation
[x] Travis

Development notes

bundle install
rake compile

https://gist.github.com/xavriley/507eff0a75d4552fa56e

Contributing

Fork it ( http://github.com//fast_osc/fork )
Create your feature branch (git checkout -b my-new-feature)
Commit your changes (git commit -am 'Add some feature')
Push to the branch (git push origin my-new-feature)
Create new Pull Request