Question

I am looking for optimisation ideas regarding the below piece of code written in Ruby. It looks for header file dependencies in a dependency file (generated by GCC using gcc -M -MM -MF)

I’ve ran the Ruby profiler and found that most of the CPU time was spent in the Array#each function below.

lines = File.readlines(depends_file)
lines.each do |line|
  if line.include?(".h") then
    line.lstrip.rstrip.chomp("\\").split(' ').each do |line|
      header_dependencies << line
    end
  end
end

The depends file contains dependencies in the makefile rule format as specified by gcc’s -M option, see GCC’s Preprocessor Options. Here’s an example from my project –

CyclicRedundancyCheck.o: src/CyclicRedundancyCheck.c \
 inc/CyclicRedundancyCheck.h inc/StdDefs.h

Can someone please post an alternate, heavily optimised solution?

Answer 1

Not heavily optimized but a bit faster here (about 20% for my test-case):

File.open(depends_file, 'r') do | io |
  io.each do |line|
    next unless line.include?('.h')
    header_dependencies.concat(line.split)
  end
end
header_dependencies.delete('\\')

I don’t think you can gain much more with Ruby itself. If this is really too slow, you should consider programming it (partly) in C (as an extension).

Note also that the approach is far from being robust. E.g.

 foo.o: src/ext.hlib/foo.c \

will wrongly be matched. Making it more robust will most likely cost performance, though.