Profile of hg clone on windows

Matt Mackall mpm at selenic.com
Mon Sep 19 15:05:43 CDT 2005


On Mon, Sep 19, 2005 at 03:47:19PM -0400, Daniel Santa Cruz wrote:
> This was ran on a Windows XP box using and IDE drive with NTFS.
> 
> I have included:
> * Size of the project (find)
> * hg clone timmings
> * hg clone -U timming
> * hg clone --profile output
> 
> This runs quite a bit slow, as you can see.  I am running python.org's
> python, and have used the module found in sourceforge for python that
> gives hardlinking capabilities.  Notice how under cygwin, even the
> find runs slow.
> 
> Daniel.
> 
> (from cygwin)
> $ time find upstream | wc
>   32410   40794 2542038
> 
> real    0m12.654s
> user    0m2.061s
> sys     0m10.826s
> 
> (now in windows' cmd)
> C:\src>hg --time clone upstream clonetimetest
> Time: real 81.354 secs (user 11.266+0.000 sys 4.875+0.000)

So we took 16 seconds to do all the crunching (user+sys), and spent
the rest of the time waiting.

> C:\src>hg --time clone -U upstream clonetimetest3
> Time: real 21.500 secs (user 4.641+0.000 sys 0.875+0.000)

And here we took 5.5 seconds seconds of CPU time.

Either your filesystem or disk is slow, or there's significant other load on
your machine.

> 
> (sorry for the clutter)
> C:\src>hg --profile clone upstream cloneprofiletest
>          374722 function calls (365151 primitive calls) in 128716.482
> CPU seconds

Huh. That looks like a day and a half. Unfortunately, the profile
won't tell us anything about the bulk of the time as it's outside the
program.

-- 
Mathematics is the supreme nostalgia of our time.


More information about the Mercurial mailing list