Extending Ruby with Rust benchmarks

Benchmarks Helix vs FFI vs Rutie when extending Ruby with Rust.

Helix and Rutie use C extensions to extend Ruby and compiles a native binary which is then loaded by Ruby.

FFI, on the other hand, uses libffi to call C from Ruby and requires dynamic shared objects (.dylib / .so).

Also, the Ruby implementation is just to have an idea of how the Rails blank? method compares to the Rust implementations. I'm totally aware this is like comparing apples and oranges as the ruby benchmark uses a regular expression.

The benchmarks are relatively basic for now but contributions are welcome.

How to run the benchmarks

Make sure ruby is installed with --enable-shared.

$ ruby-install ruby 2.6.1 -- --enable-shared

$ bundle install
$ bundle exec rake bench

Results

Warming up --------------------------------------
 helix - blank check   114.597k i/100ms
   ffi - blank check   219.232k i/100ms
 rutie - blank check   216.659k i/100ms
  ruby - blank check   152.847k i/100ms
Calculating -------------------------------------
 helix - blank check      1.553M (± 6.2%) i/s -      7.793M in   5.042093s
   ffi - blank check      4.230M (± 4.6%) i/s -     21.266M in   5.038519s
 rutie - blank check      4.274M (± 2.7%) i/s -     21.449M in   5.023018s
  ruby - blank check      2.316M (± 2.0%) i/s -     11.616M in   5.017375s

Comparison:
 rutie - blank check:  4273585.4 i/s
   ffi - blank check:  4230487.4 i/s - same-ish: difference falls within error
  ruby - blank check:  2316146.4 i/s - 1.85x  slower
 helix - blank check:  1552843.9 i/s - 2.75x  slower