[ODE] Comparison......

Sergio Valverde svalverde at barcelona.ubisoft.es
Thu Aug 7 03:09:01 2003


The main obstacle is avoid manipulation of <huge> matrices (like in the
original Russ' approach).. That is, the core of the ODE constraint solver. 
With the PS2 the important issue is realize how to squeeze more temporal and
spatial coherence from your source code because the <microscopic> data &
instruction caches.  The PS2 is not a general purpose computer but a very
specialized bunch of hardware. Memory fragmentation is a relatively
high-level issue. What is important in the PS2 is to localize the low-level
core loops and adapt them in a way that exploits the peculiarities of this
machine (that is, scratchpad-ram , fast dma transfers and Vus)

-----Original Message-----
From: Aras Pranckevicius [mailto:nearaz@interamotion.com] 
Sent: jueves, 07 de agosto de 2003 11:17
To: ode@q12.org
Subject: Re: [ODE] Comparison......

> Things get simpler if you start from the iterative solver (aka
> 'stepfast')... like we already did for our next PS2 game.

Have you had problems with memory fragmentation, or fitting ODE into VUs,
for example?


Aras Pranckevicius aka NeARAZ
http://www.gim.ktu.lt/nesnausk/nearaz/

_______________________________________________
ODE mailing list
ODE@q12.org
http://q12.org/mailman/listinfo/ode