From: Tomas Brixi <tomas_brixi@...>
Date: 2002-02-24T04:32:04+09:00
Subject: Digest classes performance

Hello,

I have have done some test (especially) with Digest classes on MS Windows. 
Due to __very poooooor__ performance of mswin version I have run the same test with
mingw and cygwin ruby version and also with python.

The following are the speed results:

SHA1
 0.296 MB/s		ruby 1.66 mswin/mingw *
18.041 MB/s		activestate python 2.1		
12.797 MB/s		ruby 1.65 cygwin - by pragmatic programmers

MD5
 0.298 MB/s		ruby 1.66 mswin/mingw *
22.158 MB/s		activestate python 2.1		
15.433 MB/s		ruby 1.65 cygwin - by pragmatic programmers


* - mswin version - 1.66 by pragmatic programmers
  - mingw version - 1.66 from http://www.ruby-lang.org/
These two ruby builds performs almost the same.

Tested on
CPU: AMD Duron 1GHz
HDD: STSeagate Baracuda IV - 40GB	
Windows XP and Windows 2000

The absolute numbers from the results are not so important as the
fact that python significantly outperforms ruby :-( 

The digest functions are written in C and so the performance should be
quite well.
Cygwin version is still usable but what about mswin/mingw?
Is it caused by poor io performance or bad memory management under mswin 
or anything else?


Regards

  Tom


These are the used functions

---- ruby source ----
def FileHash( fName )
	hash = Digest::SHA1.new
	if not FileTest.file?( fName )
		return ""
	end
	
	inF = File.new( fName, "rb" )
	bytesRead = 0
	bytesTotal = File.size( fName )
	while true
		buf = inF.read( 32768 )
		if buf == nil
			break
		end
		hash.update( buf )
		bytesRead += buf.length
		if fName.length > 40
			fName = fName[0..16] + "~~~~~" + fName[-16..-1]
		end
		$stdout.write( "\r%-40s :  %d of %d" % [ fName, bytesRead, bytesTotal ] )
	end
	return hash.hexdigest
end
--------

---- python source ----
def FileHash( fName ):
	if not os.path.isfile( fName ):
		return ""
	dig = sha.sha()
	inF = open( fName, "rb" )
	bytesRead = 0
	bytesTotal = os.path.getsize( fName )
	while 1:
		buf =  inF.read( 32768 )
		if buf == "":
			break
		dig.update( buf )
		bytesRead = bytesRead + len( buf )
		sys.stdout.write( "\r%s :  %d of %d" % ( fName, bytesRead, bytesTotal ) )
	inF.close()
	return dig.hexdigest()
--------



__________________________________________________
Do You Yahoo!?
Yahoo! Sports - Coverage of the 2002 Olympic Games
http://sports.yahoo.com