> I should point out that the strategy for multiplication in Tom and
> Robert's paper http://arxiv.org/abs/0901.1413 is likely to be better.
> Judging from the timings in that paper we are about a factor of two
> behind them. I plan to implement/port their very cool trick for finite
> extension fields at some point in the future. The trick is limited to
> multiplication as far as I understand it thus it might make still sense
> to consider my representation for the elimination base case etc.

Our runtime is something like O(e^1.6 n^2.8) using Karatsuba on the
matrix slices.  Looks like you've implemented a lookup table for your
finite field arithmetic?  If that's only twice as slow as the
bitslicing strategy, I'm very impressed -- especially since you have
to use 16 bits to store a 9-bit field element due to alignment.  I
guess the real question is, how quickly can one transition between the
two representations?

> Anyway ... question: Do we want my new code in Sage?

+1

-- 
To post to this group, send an email to [email protected]
To unsubscribe from this group, send an email to 
[email protected]
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

Reply via email to