Microbenchmark showing the performance penalty of simply specifying a block capture i method arguments

block_capture_penalty.rb

#!/usr/bin/env ruby

require 'benchmark'

N = 10_000_000

def no_capture
end

def block_capture(&_block)
end

Benchmark.bmbm do |b|
  b.report('nop')                             { N.times{ } }
  b.report('no capture no block passed')      { N.times{ no_capture } }
  b.report('block capture no block passed')   { N.times{ block_capture } }
  b.report('no capture with block passed')    { N.times{ no_capture {} } }
  b.report('block capture with block passed') { N.times{ block_capture {} } }
end

__END__

$ ./block_capture_penalty.rb
Rehearsal -------------------------------------------------------------------
nop                               0.430000   0.000000   0.430000 (  0.433004)
no capture no block passed        0.690000   0.000000   0.690000 (  0.696997)
block capture no block passed     0.850000   0.000000   0.850000 (  0.845417)
no capture with block passed      0.760000   0.010000   0.770000 (  0.762127)
block capture with block passed   5.230000   0.020000   5.250000 (  5.268700)
---------------------------------------------------------- total: 7.990000sec

                                      user     system      total        real
nop                               0.420000   0.000000   0.420000 (  0.423946)
no capture no block passed        0.690000   0.000000   0.690000 (  0.691389)
block capture no block passed     0.820000   0.000000   0.820000 (  0.823047)
no capture with block passed      0.720000   0.000000   0.720000 (  0.714247)
block capture with block passed   5.170000   0.020000   5.190000 (  5.194559)

$ ruby -e 'def r(a,b); (a-0.42)/(b-0.42); end; puts r(0.820000,0.690000), r(5.170000,0.720000)'
1.4814814814814816
15.833333333333334

$ ruby -e 'puts RUBY_PLATFORM,RUBY_VERSION'
x86_64-darwin13
2.2.0