On 12/27/2010 5:15 PM, Stavros Macrakis wrote:
> You're absolutely right -- I should have been looking at the max, not
> the sum, of the individual times. Very stupid of me.
>
> -s
>
Yeah, max, not sum would be right. But here's a question. Why, if we
are running on a single processor, is it only 50%
busy? Is it 50% = 100% X (1/2) because it is one processor out of
two? Or is it 50% of one processor and 0% on
the other processor?
If busy-ness is on a per processor basis,
why should any processor be less than 100% busy unless it is waiting for
disk or memory or other I/O.?
And for making a parallel matrix multiply, it seems to me that using
someone else's routine is the way to go,
not to use lisp threads. There is parallel LAPACK
www.netlib.org/sca*lapack*/