Re: performance issue with cgywin make

[Bill Hoffman]

Earnie Boyd wrote:
> Quoting Bill Hoffman <address@hidden>:
>
> > real    0m39.404s
> > user    0m0.031s
> > sys     0m0.062s
>
> Hmm...  Are these from Cygwin's shell?  Is it possible that these 
> timings do not reflect actualness due to your native program would not 
> use the Cygwin runtime?  Are you comparing apples to apples?  You 
> might want to find some real measuring software.  I've also noticed 
> that times can be skewed by the caching within the drive controller as 
> well.
I ran the tests several times, and the timings seemed to be pretty 
stable.   Also, it was mostly stat'ing the same files
over and over again.  So, things should have been very cached for the 
most part.  The numbers are in line
with what we found with cmake about a year ago.  At that point we were 
using a profiler.  I was running this
from the a cygwin shell, and I compiled the program both with and 
without the cygwin compiler.   I know with
cmake we saw at least a 2 fold speed up when we went from _stat to the 
native windows calls.

-Bill