[Open-graphics] [OT?] Fwd: Direction of graphics card design

[Jared Putnam]

(Eh...James, did you intend your reply to be off list?  This can't be
any more off topic than the last big thread...Anyway, I've extracted
the only really on-topic part.)

--- James Richard Tyrer <tyrerj at acm.org> wrote:
> And, AFAIK, fixed point only exists in software.

There are DSPs that do it.

> My somewhat excessive design spec was for registers which held 4
32bit 
> floats and 4 hardware floating point multipliers that could be used
with 
> FMACs to do a [4,4] * [4,4] matrix multiply in 4 clocks and a four
input 
> floating point adder to do [4,4] x [4,4] matrix multiply in 16
clocks.

ATi seems to think SIMD is the right approach[1].  These big
multipliers certainly look excessive, though.  The only way to know
what architecture to go with is to get statistics on how often
individual instructions are executed.  Actually, if OGA built profiling
counters into the chip, would that be useful to programmers?  Even if
the performance wasn't competitive, would developers still buy the chip
for the feedback?


[1]http://enthusiast.hardocp.com/article.html?art=ODIyLCwsaGVudGh1c2lhc3Q=


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com