Re: RENDER ext. & hardware acceleration, libart integration

Around 19 o'clock on Aug 26, Juliusz Chroboczek wrote:

> Wouldn't the resulting cache pollution be really bad?  (Shadowfb
> suffers from the same problem.)

It wouldn't be as bad as shadowfb -- any operations done by the card 
wouldn't affect the cache contents; the cache would only be filled with 
data that the CPU was drawing.  The more acceleration you could provide, 
the less polution would occur.

I think it would be worth measuring the performance effect of caching the 
frame buffer on frame buffer reads.  If the CPU can burst traffic across 
the AGP in both directions, we may see a significant improvement.  Effects 
on the rest of the system could then be judged with respect to any 
possible graphics performance improvements.

Keith Packard        XFree86 Core Team        HP Cambridge Research Lab