This document provides a quick guide to profiling Rails applications with kcachegrind/qcachegrind in 3 steps: 1) Install kcachegrind or qcachegrind, 2) Add a development_profiler.rb file to abstract profiling configuration, 3) Wrap code in .prof blocks to profile specific sections and view results in qcachegrind. It includes a contrived example of profiling inefficient code that initializes an array of symbols and an improved version.

1. Quick guide to CPU
profiling Rails apps with
k/qcachegrind
Mike Auclair
@mikeauclair
github.com/mikeauclair

2. Step 1:
Acquire kcachegrind(KDE) or qcachegrind
(everything else)
● qcachegrind is available in homebrew

3. Step 2:
Grab development_profiler.rb or something
similar to asbtract away some of the profiling
configuration, plop it in lib/

4. development_profiler.rb
class DevelopmentProfiler
def self.prof(file_name)
if Rails.env.development?
RubyProf.start
yield
results = RubyProf.stop
File.open "#{Rails.root}/tmp/performance/callgrind.#{file_name}.clt", 'w' do |file|
RubyProf::CallTreePrinter.new(results).print(file)
end
else
yield
end
end
end

5. Step 3:
Wrap your suspect code in a .prof block:
DevelopmentProfiler.prof('junk') do
SlowThing.new(45000)
end

6. Step 4:
Run your code and bust open the file in
qcachegrind

7. Contrived Example
class Shitty
def initialize(some_int)
@symbols = (0..some_int).map(&:to_s).map(&:to_sym)
end
end

9. Improved Contrived Example
class Better
def initialize(some_int)
@symbols = (0..some_int).map{|i| i.to_s.to_sym}
end
end