Adventures in Optimization.. When a Ruby primitive is the Hog


#1

So you’ve run your code with -rprofile or require ‘profile’

And the answer was IO.each_line

Which is a bit like saying the answer to the Question of Life, the
Universe and Everything is “42”.

a) You’ve used IO.each_line in many places.
b) There is nothing you can do to speed it up.

Enter Carter’s Canny Primitive Profiler…

$f=Hash.new(0)

CUNNING_MIN_CARE = 1 # Only care about stack frames that have been
called
# at least CUNNING_MIN_CARE times
CUNNING_DEPTH = 2 # Profile base on CUNNING_DEPTH number of
# stack frames of this calls that that
calls… that
# calls Array.each
# Make it -1 if you want the lot.

at_exit{
$f.keys.find_all{|k| $f[k] >= CUNNING_MIN_CARE}.sort_by{|k|
$f[k]}.each{|k|
puts “\n\n#{k} >>>#{$f[k]}<<<”
}
}

class IO
alias_method :orig_each_line, :each_line
def each_line(&block)
i=0
orig_each_line do |l|
block.call(l)
i+= 1
end
ensure
$f[caller(1)[0…CUNNING_DEPTH].join("\n")] += i
end
end

On exit you can see which invocation of each_line via which path was
the busiest!

Of course, sometimes we don’t actually care about CPU time. The by
several orders of magnitude, the slowest operation in a modern PC is
pulling stuff of the disk.

So how about this variation on the Theme…
class IO
alias_method :orig_read, :read
def read(*arg)
start = Time.now
orig_read(*arg)
ensure
$f[caller(1)[0…CUNNING_DEPTH].join("\n")] += Time.now - start
end
end

I love Ruby :slight_smile:

John C. Phone : (64)(3) 358 6639
Tait Electronics Fax : (64)(3) 359 4632
PO Box 1645 Christchurch Email : removed_email_address@domain.invalid
New Zealand