On Dec 10, 2007 9:57 AM,  <[EMAIL PROTECTED]> wrote:

> Fundimentally, it's a 8-way vector unit.  But rather than the vectors being 
> math ops, they're memory blocks.  In addition to the normal block-moves, it 
> also has pattern fill, overlay, barrel shifting, and merging.  Setup time 
> should be 3 cycles, and once running can do very large units of memory 
> control in a limited time.  In my design the GPU had no direct access to RAM, 
> only to a small internal fifo buffer, which the blitter would continuously 
> feed in and out.  This kept the GPU itself very small.
>

Are you able to keep the memory controller saturated?  For 2D, I had
planned on a fifo-based engine with just bitblt and solid fill for
rectangles.  Perhaps a different architecture could get better
performance with less logic.

How many blocks does the processor have access to at once, how wide
are they, and what fills and empties them?


-- 
Timothy Normand Miller
http://www.cse.ohio-state.edu/~millerti
Open Graphics Project
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to