Re: display list optimization / performance
- From: Wolfgang Draxinger <wdraxinger@xxxxxxxxxxxxxxxx>
- Date: Sun, 20 Jan 2008 17:40:41 +0100
Paul Geisler wrote:
Hello.
I use hierarchical rendering, lets say a cube, used in many
instances, maybe different materials but most of the time
diferent size or position. I thought that putting a basic cube
to a display list, then use it in several calls preceeded
transformations would be fine. But it isn't. Even when i
compile a bunch of transforms and cube-calls to a display list,
the performance is bad compared to a single quads containing
many cubes. It seems the display list does nothing as calling
all theire statements quite eficcient. Are there no further
optimizations possible? Having T&L in hardware it seems weird
that filling a poly array with cpu-transformed instances is the
way to go :-(
The ratio between actually rendered faces and the overhead of
creating a displaylist vs. rendering it in immediate mode is
largely biased to the overhead.
Instead of simple cubes try really complex geometry. Then you can
actually gain the performance of display lists.
Wolfgang Draxinger
--
E-Mail address works, Jabber: hexarith@xxxxxxxxxx, ICQ: 134682867
.
- Follow-Ups:
- Re: display list optimization / performance
- From: Paul Geisler
- Re: display list optimization / performance
- References:
- display list optimization / performance
- From: Paul Geisler
- display list optimization / performance
- Prev by Date: display list optimization / performance
- Next by Date: Re: framebuffer object
- Previous by thread: display list optimization / performance
- Next by thread: Re: display list optimization / performance
- Index(es):
Relevant Pages
|
Loading