Sharon Clay (src++at++rose)
Tue, 27 Feb 1996 09:23:54 -0800
It is an RE implementation issue. The IRIS GL implementation had these
limitations on RE and the OpenGL implementation for RE has the same
basic functionality.
->- if this bug will be corrected in the future on these machines,
Absolutely - IMPACT and InfiniteReality do not have these limitations.
->
->If this extension works without bugs on Infinite Reality, it would be
->interesting to know a little more about subload performances :
->- what is the absolute cost (in ms) for the subload of a non-mipmapped and a
->mip-mapped texture with a given size ?
IR texture downloads are several times faster than RE and IMPACT and is
initially host limited. You should initially see 160+++MBytes/sec,
depending on CPU (R4K, R10K, ...).
With these kinds of rates, the real issue will be how fast you
can get the data off disk.
->- how does the texture size modify this cost ?
Bigger is always faster but I think you can do quite well wtih small
loads < 32x32.
->- what is the best place in the DRAW process to subload textures ?
At the end of the frame, loading them for the next frame.
This is because the IR has parallelism in the pipe that lets you fill
texture memory as you are doing rendering so you can overlap the
fill-limited time at the end of the frame with setting off the texture
downloads. Then, when you are ready to use the textures in the next frame
they will be ready. Doing it at the start of the frame will delay your
drawing and doing it on-demand wil cause interruptions
in the drawing stream.
src.
--
-----{-----{---++at++ -----{----{---++at++ -----{----{---++at++ -----{----{---++at++
Sharon Rose Clay (Fischler) - Silicon Graphics, Advanced Systems Dev.
src++at++sgi.com (415) 933 - 1002 FAX: (415) 965 - 2658 MS 8U-590
-----{-----{---++at++ -----{----{---++at++ -----{----{---++at++ -----{----{---++at++
This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:52:27 PDT