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]

Re: higher-level IO very slow with cygwin1.dll 5.10 (due to set_flags?)


On Sat, Jun 26, 2004 at 02:36:52AM -0400, Joseph wrote:
> Thanks to those who responded to my post.
> 
> I am not sure though if all of you read the
> entire sequence of the thread.
> 
> In my first post, I did attach an excerpt
> from an strace dump of a sample program that
> seemed to show "where the time is going".

You initial accusation against set_flags was unfounded
and led nowhere. But you have a very good point.

Your "fast" program is doing reads in blocks of 1024.
Your "slow" program is using getc.

I observed that the slow program was 20 times slower
than the fast one, even though strace showed that getc
was properly calling read in blocks of 1024.

I then noticed that getc is not implemented as a macro
on Cygwin. It's a newlib function
 
74      int
75      _DEFUN(getc, (fp),
76             register FILE *fp)
77      {
78        int result;
79        _flockfile (fp);
80        /* CHECK_INIT is called (eventually) by __srefill.  */
81        result = __sgetc (fp);
82        _funlockfile (fp);
83        return result;
(gdb) 
84      }

_flockfile calls a bunch of pthread functions. That's where all
the time is spent, a large part of it verifying that pointers
are valid.

It's probably a relatively new feature of Cygwin. Others may know
more about the history and justification of implementation.

Pierre

--
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]