A little bird told me that FD_ZERO() burns lots of cycles in
apr_wait_for_io_or_timeout(). It turns out that this is an easy
conversion to poll(), which doesn't have such overhead in the
interface.
This works for me with some testing (timeouts on read and write work
for me).
--- /tmp/sendrecv.c
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
>
> A little bird told me that FD_ZERO() burns lots of cycles in
> apr_wait_for_io_or_timeout(). It turns out that this is an easy
> conversion to poll(), which doesn't have such overhead in the
> interface.
>
> This works for me with some test
"Ryan Bloom" <[EMAIL PROTECTED]> writes:
> > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
> >
> > A little bird told me that FD_ZERO() burns lots of cycles in
> > apr_wait_for_io_or_timeout(). It turns out that this is an easy
> > conversion to poll(), which doesn't have such overhead in th
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
>
> "Ryan Bloom" <[EMAIL PROTECTED]> writes:
>
> > > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
> > >
> > > A little bird told me that FD_ZERO() burns lots of cycles in
> > > apr_wait_for_io_or_timeout(). It turns out that this is an eas
> > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
> >
> > A little bird told me that FD_ZERO() burns lots of cycles in
> > apr_wait_for_io_or_timeout(). It turns out that this is an easy
> > conversion to poll(), which doesn't have such overhead in the
> > interface.
> >
> > This works for me
> From: Bill Stoddard [mailto:[EMAIL PROTECTED]
>
>
> > > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
> > >
> > > A little bird told me that FD_ZERO() burns lots of cycles in
> > > apr_wait_for_io_or_timeout(). It turns out that this is an easy
> > > conversion to poll(), which doesn't hav
> > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
> >
> > "Ryan Bloom" <[EMAIL PROTECTED]> writes:
> >
> > > > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
> > > >
> > > > A little bird told me that FD_ZERO() burns lots of cycles in
> > > > apr_wait_for_io_or_timeout(). It turns out that
> > From: Bill Stoddard [mailto:[EMAIL PROTECTED]
> >
> >
> > > > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
> > > >
> > > > A little bird told me that FD_ZERO() burns lots of cycles in
> > > > apr_wait_for_io_or_timeout(). It turns out that this is an easy
> > > > conversion to poll(), wh
"Ryan Bloom" <[EMAIL PROTECTED]> writes:
> > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
> >
> > "Ryan Bloom" <[EMAIL PROTECTED]> writes:
> >
> > > > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
> > > >
> > > > A little bird told me that FD_ZERO() burns lots of cycles in
> > > > apr_w
> From: Bill Stoddard [mailto:[EMAIL PROTECTED]
>
>
> > > From: Bill Stoddard [mailto:[EMAIL PROTECTED]
> > >
> > >
> > > > > From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED]
> > > > >
> > > > > A little bird told me that FD_ZERO() burns lots of cycles in
> > > > > apr_wait_for_io_or_timeout().
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
>
> "Ryan Bloom" <[EMAIL PROTECTED]> writes:
>
> > > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
> > >
> > > "Ryan Bloom" <[EMAIL PROTECTED]> writes:
> > >
> > > > > From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED]
> > > > >
> > > > > A l
Ryan Bloom wrote:
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
"Ryan Bloom" <[EMAIL PROTECTED]> writes:
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
"Ryan Bloom" <[EMAIL PROTECTED]> writes:
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED]
A little bird told
> From: Brian Pane [mailto:[EMAIL PROTECTED]
>
> Ryan Bloom wrote:
>
> >>From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
> >>
> >>"Ryan Bloom" <[EMAIL PROTECTED]> writes:
> >>
> >>
> >>
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
>
> "Ryan Bloom" <[EMAIL PROTECTED]> writes:
> Have either of you benchmarked with apr_poll() or are you assuming that
> the problem exists?
>
> Ryan
>
>
Sorry didn't answer you here... There definitely are extra instructions and
function calls involved with using apr_poll() in this case. I don't know the
exact number but I could find out.
> From: Brian Pane [mailto:[EMAIL PROTECTED]
> Sent: 03 July 2002 18:05
>>You are missing the point. If apr_poll() is to be useful to external
>>projects, then it must perform well. If it performs so poorly that we
>>refuse to use it inside of APR, then it couldn't possibly be useful to
>>extern
"Bill Stoddard" <[EMAIL PROTECTED]> writes:
> > Have either of you benchmarked with apr_poll() or are you assuming that
> > the problem exists?
> >
> > Ryan
> >
> >
>
> Sorry didn't answer you here... There definitely are extra instructions and
> function calls involved with using apr_poll() in t
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED] Behalf Of Jeff Trawick
> Sent: 03 July 2002 21:16
> "Bill Stoddard" <[EMAIL PROTECTED]> writes:
>
> > > Have either of you benchmarked with apr_poll() or are you assuming that
> > > the problem exists?
> > >
> > > Ryan
> > >
> > >
> >
> > Sor
Bill Stoddard wrote:
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
I'd rather pay the maintenance overhead on this code. BTW, testing on AIX
yields up to a 9% increase in throughput with this patch (serving 500 byte
file out of mod_mem_cache). Huge win for a small change!
Bill
fantastic speed
>
> On Wed, 3 Jul 2002, Ian Holsman wrote:
>
> > Bill Stoddard wrote:
> > >>>From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
> >
> >
> >
> > > I'd rather pay the maintenance overhead on this code. BTW,
> testing on AIX
> > > yields up to a 9% increase in throughput with this patch
> (serving 500
[EMAIL PROTECTED] wrote:
Here is the patch to use apr_poll instead of select. This is just being
posted for completeness, not because I believe it will actually perform as
well as the standard poll() implementation. Bill offered to benchmark it,
so I am posting it. I will re-implement apr_poll()
On Wed, 3 Jul 2002, Ian Holsman wrote:
> Bill Stoddard wrote:
> >>>From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
>
>
>
> > I'd rather pay the maintenance overhead on this code. BTW, testing on AIX
> > yields up to a 9% increase in throughput with this patch (serving 500 byte
> > file out of
Here is the patch to use apr_poll instead of select. This is just being
posted for completeness, not because I believe it will actually perform as
well as the standard poll() implementation. Bill offered to benchmark it,
so I am posting it. I will re-implement apr_poll() tomorrow to offer much
On Fri, 5 Jul 2002, Ian Holsman wrote:
> [EMAIL PROTECTED] wrote:
> > Here is the patch to use apr_poll instead of select. This is just being
> > posted for completeness, not because I believe it will actually perform as
> > well as the standard poll() implementation. Bill offered to benchmark i
If that was all we were doing, I would agree with you. But `Jeff's patch
implements BOTH select and poll with an #ifdef, because not every platform
has poll(). This is exactly the reason for having apr_poll(), and not
using it is stupid. If the argument is performance, then back it up with
numbe
On Fri, 5 Jul 2002, Roy T. Fielding wrote:
> > If that was all we were doing, I would agree with you. But `Jeff's patch
> > implements BOTH select and poll with an #ifdef, because not every platform
> > has poll(). This is exactly the reason for having apr_poll(), and not
> > using it is stupid.
As promised, a new implementation of apr_poll, which should improve
performance. I have only implemented the HAVE_POLL case so far, but the
API is 99% the same, so I have no fear that the others are not
possible. In fact, the ONLY change to the function API, is that
apr_poll() itself takes one m
Oh, one more advantage to this code, is that most of the
apr_poll() support functions can either go away, or they can be merged to
a common implementation. The only function that won't be common is
apr_poll() itself.
Ryan
[EMAIL PROTECTED] wrote:
As promised, a new implementation of apr_poll, which should improve
performance. I have only implemented the HAVE_POLL case so far, but the
API is 99% the same, so I have no fear that the others are not
possible. In fact, the ONLY change to the function API, is that
apr_p
> From: Brian Pane [mailto:[EMAIL PROTECTED]
>
> [EMAIL PROTECTED] wrote:
>
> >As promised, a new implementation of apr_poll, which should improve
> >performance. I have only implemented the HAVE_POLL case so far, but
the
> >API is 99% the same, so I have no fear that the others are not
> >possib
On Fri, Jul 05, 2002 at 08:47:32PM -0400, [EMAIL PROTECTED] wrote:
> This also creates a support library for APR, this is basically just a
> series of functions that APR can use internally to get the job
> done. Since wait_for_io_or_timeout was identical between files and
> sockets, I have moved t
> From: Justin Erenkrantz [mailto:[EMAIL PROTECTED]
>
> On Fri, Jul 05, 2002 at 08:47:32PM -0400, [EMAIL PROTECTED] wrote:
> > This also creates a support library for APR, this is basically just
a
> > series of functions that APR can use internally to get the job
> > done. Since wait_for_io_or_ti
On Sat, Jul 06, 2002 at 08:32:18AM -0700, Ryan Bloom wrote:
> Because if Apache can't get it right, then I am assuming that nobody
> else can either. I had originally coded it to use *nsds just as you
> describe below, and it didn't pass any tests, because throughout the
> code people were passing
> From: Justin Erenkrantz [mailto:[EMAIL PROTECTED]
>
> On Sat, Jul 06, 2002 at 08:32:18AM -0700, Ryan Bloom wrote:
> > Because if Apache can't get it right, then I am assuming that nobody
> > else can either. I had originally coded it to use *nsds just as you
> > describe below, and it didn't pa
On Sat, Jul 06, 2002 at 12:11:59PM -0700, Ryan Bloom wrote:
> I vote to fix the API so that these kinds of mistakes can't happen in
> the future. I made a lot of mistakes when I designed APR (even though
> Manoj tried to convince me I was wrong). One of those mistakes is
> having functions use a
> From: Justin Erenkrantz [mailto:[EMAIL PROTECTED]
>
> On Sat, Jul 06, 2002 at 12:11:59PM -0700, Ryan Bloom wrote:
> > I vote to fix the API so that these kinds of mistakes can't happen
in
> > the future. I made a lot of mistakes when I designed APR (even
though
> > Manoj tried to convince me I
On Sat, Jul 06, 2002 at 12:11:59PM -0700, Ryan Bloom wrote:
> I vote to fix the API so that these kinds of mistakes can't happen in
> the future.
+1, let's fix the API now. (I also strongly dislike input/output parameters.)
-aaron
At 11:47 AM 7/7/2002, Aaron Bannert wrote:
On Sat, Jul 06, 2002 at 12:11:59PM -0700, Ryan Bloom wrote:
> I vote to fix the API so that these kinds of mistakes can't happen in
> the future.
+1, let's fix the API now. (I also strongly dislike input/output parameters.)
Microsoft's Win32 API was built
37 matches
Mail list logo
