Re: Poor performance on 64-bit Windows x86_64 systems

[Mike McCool <mlm at aechelon dot com>]

tom honermann wrote:
> I'm new to this mailing list but have seen several posts in the mail 
> archives regarding Cygwin's poor performance on 64-bit Windows/x86_64 
> OSs.  The last post I saw was from David Morgan in late January 
> indicating he was working on the problem.  Any updates?  Any hope of a 
> solution before 1.7.2 is released?  I may be able to help track down the 
> problem, test, or code a solution, but I'd like to know what has been 
> discovered so far before I jump in.

Dave and I spent a while trying to track this down before getting
distracted into other things.  Unfortunately, we didn't make much
progress except to verify that the problem was somehow introduced
between 1.5.19-4 and 1.5.20-1.  We discovered based on a find test
vs. a simpler date loop test that there may have been two slowdowns,
the second introduced sometime during 1.5.20-1 that only seemed to
affect the find test.  The date loop test showed a slowdown between
1.5.19-4 and 1.5.20-1.  The tests were, basically, how long does
it take to find *.cpp in the current directory, and how long does
it take to run date N thousand times.

We used a binary search in CVS to try and pinpoint the problem but
were unable to find a specific change.  Further, a 1.5.19-4
cygwin1.dll built with the build environment from 1.5.20-1
produced a slow build.  That made us think it might even have
been a gcc upgrade that happened in that time, and it may be
gcc causing the issue.

However, that was about when we both got devoured in other
activities, so we haven't had a chance to pick it back up again.
(Possibly Dave has and he hasn't mentioned it to me.)  I've been
meaning to get back to it soon.
	--MLM