Re: Benefits of Display listing of geosets

New Message Reply Date view Thread view Subject view Author view

Angus Dorbie (dorbie++at++bitch.reading.sgi.com)
Fri, 20 Sep 1996 10:05:20 +0100


The funky DMA pull mode also works on iR with patch 1355.
Even with increased bandwidth I wouldn't expect to see
immediate mode match the performance of DMA on iR unless
you use packed vertex arrays which should also ultimately
use graphics DMA too.

My 2p.

Angus.

On Sep 19, 3:57pm, Guy Russell wrote:
> Subject: Re: Benefits of Display listing of geosets
> On Sep 16, 6:07pm, Michael T. Jones wrote:
> > Subject: Re: Benefits of Display listing of geosets
> > On Sep 16, 12:43pm, Ken Sakai wrote:
> > > Subject: Benefits of Display listing of geosets
> > >
> > > Perfomers,
> > >
> > > Does anyone have an idea of how much of a performance increase
> > > pfGSetDrawMode(geoset, PFGS_COMPILE_GL, val) will tend to buy you
> > > on the various platforms? I know that objects drawn using OpenGL
> > > display lists display MUCH faster than those drawn in immediate mode
> > > in iR's. Will OpenGL display lists buy you much on Impact graphics,
> > > RE graphics, and on Extreme graphics systems? I was told that
> > > OpenGL display lists should also run faster on Impact machines but I do
not
> > > seem to be getting any speed increase with
> > > pfGSetDrawMode(geoset, PFGS_COMPILE_GL, val) on Impacts.
> >
> > Performer's immediate-mode rendering has always been as fast or
> > faster (usually faster) than IRIS GL or OpenGL display-list mode.
> > This is true on iR also, but there's more to the story...
> >
> > ...InfiniteReality has a 15+ megabyte cache on the geometry engine
> > board that is used to keep local copies of certain display lists.
> > Not all display lists can be kept locally (i.e., those that do
> > texture binds prevent lists from being cached). For those that are
> > in fact able to be cached, the difference between cached display
> > list and immediate mode over the Onyx-to-iR bus is significant and
> > will almost always result in better performance. This is due to
> > the cache's greater bandwidth than the bus interface.
> >
> > For this reason, we made the perfly in Performer 2.1 default to the
> > display-list mode. To see the difference, use the "d" (for display
> > list) and "i" (for immediate mode) keys to toggle between the two
> > cases.
> >
> > On other machines there is not the same advantage for display lists
> > that you'll see on iR. In the future, should bus speeds increase,
> > we'll see the advantage of display list mode decrease, perhaps to
> > the point where it won't be used.
>
> It is true that other machines do not have display list memory in hardware,
> however display lists are still a potential performance win on the Impact.
 If
> a display list is pullable, the graphics will initiate the DMA for the
display
> list from main memory. Pullability keeps the processor from getting involved
> in the I/O for the display list. On a single processor system this can be a
> big win if you are CPU limited. Most vis sim oriented display lists will be
> pullable. The one major exception to date is anytime a display list binds a
> texture. This limitation will be removed in upcoming releases, but for now
you
> must keep bind texture calls out of pullable display lists.
>
> >
> > Michael
> > =======================================================================
> > List Archives, FAQ, FTP: http://www.sgi.com/Technology/Performer/
> > Submissions: info-performer++at++sgi.com
> > Admin. requests: info-performer-request++at++sgi.com
> >-- End of excerpt from Michael T. Jones
>
>
> Guy
>
> --
> Guy Russell
> Vis Sim Applied Engineering Silicon Graphics Inc.
> Email:guyr++at++engr.sgi.com 2011 N. Shoreline Blvd. M/S 3L-916
> Phone:(415)933-6113 Mountain View CA 94043
> =======================================================================
> List Archives, FAQ, FTP: http://www.sgi.com/Technology/Performer/
> Submissions: info-performer++at++sgi.com
> Admin. requests: info-performer-request++at++sgi.com
>-- End of excerpt from Guy Russell

=======================================================================
List Archives, FAQ, FTP: http://www.sgi.com/Technology/Performer/
            Submissions: info-performer++at++sgi.com
        Admin. requests: info-performer-request++at++sgi.com


New Message Reply Date view Thread view Subject view Author view

This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:53:35 PDT

This message has been cleansed for anti-spam protection. Replace '++at++' in any mail addresses with the '@' symbol.