Tomas Toft <[EMAIL PROTECTED]> writes:

> Martin Geisler wrote:
>> Like you, I had expected GF256 to be significantly faster.
>
> I don't like the fixed input data in the timings.
>
> The Zp elements chosen may be good or bad candidates, and
> computation on random elements may be worse...

That is true.

> Regarding GF(256) this should not be a problem, as IIRC the
> multiplication is a table lookup. However, you may avoid
> cache-misses entirely, so those numbers should also be taken with a
> small grain of salt.

I don't know if the Python code is small enough to be affected by
stuff like cache-misses... But I have speed up the GF256 computations
somewhat by doing just one table lookup instead of three (the code
used to use logarithm and anti-logarithm tables, it now simply uses a
multiplication table).

The time per multiplication went from 6.4 usec to 3.3 usec between
d7d3f6ddb39b and 577c1e069d06. The time for a multiplication done on
Shares is then even more dependent on the network than before.

>> I wont really have time before I return from Switzerland in
>> September (I leave in a week), but can I find the article online? I
>> found the conference webpage, but it does not link to your article,
>> and neither does your own publication list.
>
> My publication list is my fault. But the paper will be available in
> the conference proceedings; to appear in LNCS I believe. I can dig
> out a copy and mail it to you if you like.

No, don't bother, I'll wait until I get back from Switzerland.

-- 
Martin Geisler
_______________________________________________
viff-devel mailing list (http://viff.dk/)
[email protected]
http://lists.viff.dk/listinfo.cgi/viff-devel-viff.dk

Reply via email to