This is the mail archive of the cygwin mailing list for the Cygwin project.
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |
Other format: | [Raw text] |
On Sat, May 28, 2005 at 11:05:23PM -0400, Christopher Faylor wrote:
On Sat, May 28, 2005 at 08:18:46PM -0400, Christopher Faylor wrote:
I have an idea about how to work around this problem but I have to think about how dangerous it might be. Basically removing the signal handling wrapper around pthread_getspecific and pthread_setspecific. That may work ok but I have to think about worst case scenarios.
I've semi-convinced myself that pthread_[gs]etspecific do not need signal protection so I've released a snapshot which turns it off:
http://cygwin.com/snapshots/
This version is about 2.7 times faster than cygwin 1.5.17 but it is still not as fast as mingw. I don't think we're going to hit mingw performance since there are still cygwin overhead issues involved.
AFAIK, this is only going to provide a speedup for this specific case. I don't think a general cygwin user is going to notice any improvement.
Gerrit, if you have a chance could you confirm or deny if this change has any effect for you?
$ C:\cygwin\bin\time cygspd.exe cygspd.dat 94.04user 0.61system 1:41.69elapsed 93%CPU (0avgtext+0avgdata 10560maxresident)k 0inputs+0outputs (667major+0minor)pagefaults 0swaps
$ C:\cygwin\bin\time cygspd.exe cygspd.dat 264.50user 0.75system 4:48.08elapsed 92%CPU (0avgtext+0avgdata 10560maxresident)k 0inputs+0outputs (667major+0minor)pagefaults 0swaps
It is about 2,8514851485148514851485148514851 times faster at my workstation.
Gerrit -- =^..^=
-- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |