From: Radomir Mech (rmech++at++engr.sgi.com)
Date: 02/21/2002 10:02:09
Hi Andreas,
I've had similar problem when playing with multi-pass rendering. Stats mesure
only
time up to the last graphics command you send down the pipeline. If the commands
are
time consuming (e.g. pixelcopy or pixeldraw of large portions of the screen or
several full-screen polygons) stats will behave like that.
I added an internal call void __pfFinishFrameStats(void) to 2.4 that calls
glFinish
and updates the stats. It is not in any header file, but it is exposed. It is
also
in 2.5. Add the call to the end of your draw function and stats should include
all graphics calls. But use the call only for testing, glFinish() costs some
precious extra time on most systems.
Radomir
Andreas Ekstrand wrote:
>
> Hi Performers!
>
> I've got a problem with a huge post-draw (or "the time spent after
> pfDraw()" as you really should call it), stretching over several
> milliseconds though I don't really do anything after pfDraw() except for
> drawing the statistics.
>
> I have attached a screen dump of my statistics. The time for drawing
> statistics should be included in the dotted line after postdraw, so that
> couldn't be the problem. I also notice the problem (30Hz instead of
> 60Hz) even when I don't draw the statistics.
>
> I'm aware of the large pre-app (or time spent before pfFrame()) and
> that's something that I'm working on. But that couldn't really be
> important in this issue, could it?
>
> Now the really wierd fact: When I measure the time spent after pfDraw()
> using pfGetTime() I only get some 0.03 ms! Furthermore, I get a huge
> pfSync() of some 26 ms.
>
> Does anyone recognize this behaviour? What am I missing here? I'd
> appreciate any comment, suggestion or solution.
>
> Regards,
> Andreas Ekstrand
>
> --
> ---------------------------------------------------------
> FDM-AE Andreas Ekstrand |E-mail: Andreas.Ekstrand++at++saab.se
> Saab AB |Phone: +46 (0)13 - 18 40 42
> SE-581 88 Linkoping |Fax: +46 (0)13 - 18 41 77
> SWEDEN |
> ---------------------------------------------------------
>
> --------------------------------------------------------------------------------
> [Image]
This archive was generated by hypermail 2b29 : Thu Feb 21 2002 - 10:03:49 PST