24 Dec
2004
24 Dec
'04
3:55 p.m.
On Fri, 24 Dec 2004, Jim Battle wrote:
re: 9 bits
We doubled the
peak throughput by adding only 12.5% more datapath resources. The
second reason was that the chip used RAMBUS memories, which were only sold
with 9b channels, and we had no particular need for parity.
That's pretty clever, a good use for those parity RAMs! Two-wide
would make 18 bits, a really nice width (3 characters).
The multipliers had various rounding modes, including
"round random", where 6b
pseudorandom bias was added just below the trucation point, as some audio
filter algorithms have better spectral properties with this type of rounding.
Man, I love bizarre hardware solutions to problems like that. It
all sounds so baroque.