delorie.com/archives/browse.cgi   search  
Mail Archives: cygwin-developers/1998/06/03/11:31:28

From: cgf AT cygnus DOT com (Christopher Faylor)
Subject: RE: Update on status of submitted patches
3 Jun 1998 11:31:28 -0700 :
Message-ID: <199806031821.OAA11699.cygnus.cygwin32.developers@kramden.cygnus.com>
To: sos AT prospect DOT com DOT ru
Cc: cygwin32-developers AT cygnus DOT com

>Guy Gascoigne - Piggford wrote:
>> At 06:18 PM 6/2/98 -0700, you wrote:
>> >3) We're going to hold off on the changes to support -pipe in gcc
>> >because gcc -pipe doesn't actually yield a performance improvement on
>> >the Win32 host.  These changes may go in eventually.
>> 
>> Out of curiosity, is there a performance improvement if you a dual CPU
>> machine?
>
>Yes there is.  Moreover, I see some improvement on a single CPU
>machine, also there is no temp files with -pipe...

Do you have figures to show the performance improvement?  As I mentioned
previously, I tried this and saw absolutely no change.  Perhaps, I patched
something wrong.  I added your changes to cygwin.din, fhandler.cc, pipe.cc,
fcntl.h, and winsup.h.  Then I removed most of the CYGWIN32 code in
pexecute.c, leaving only a define for fix_argv and a pwait function.

Prior to my change I got an internal compiler error if I tried to use -pipe.
After the change I got no error but I did not notice any improvement at
all.

To test, I compiled several different groups of files in winsup.  The results
were consistent.

If I missed something, please let me know.  Possibly, as I mentioned previously
my hardware is good enough that a speed improvement would be masked.

cgf

- Raw text -


  webmaster     delorie software   privacy  
  Copyright 2019   by DJ Delorie     Updated Jul 2019