hutch@net1.ucsd.edu (Jim Hutchison) (04/17/88)
<>
The company at which I am currently gainfully employed is moving it's
software and "new" boards to a new platform, a VME (2) based system.
In all probability this will be a SUN 3 or 4, although Apollo does keep
being mentioned. This is the background, not the question.
For fast operations of a graphics processor on a large (4k x 4k) frame
buffer, it seems logical that one would need a fast private bus with
which to access the frame buffer. The operations in question are such
things as NxN convolutions, hsv change, filters, masks, warps, etc. for
which the processor(s) can not fit on the board with the frame buffer
bric-a-brac.
At what point does a 12.5 MHz VME (2), with a "normal" process mix running
on it, not have enough left over bandwidth to handle drawing on the
frame buffer? This would appear to be a sliding scale based on the processing
power of the graphics processor and the access time for the frame buffer.
The "burst" rate certainly figures into this for things like uploading rasters,
presuming that the frame buffer & graphics processor will be able to deal
with it.
Any help much appreciated.
Jim Hutchison UUCP: {dcdwest,ucbvax}!cs!net1!hutch
ARPA: Hutch@net1.ucsd.edu
Disclaimer: I'm dead, really, deceased, this is a recording.