GMP 4.3 multiplication performance

[Torbjorn Granlund]

bodrato at mail.dm.unipi.it writes:

  > But before doing more work on mpn_mul, I want more toomMN primitives, in
  > particular toom43, toom53 (which exists but is currently unused) and
  
  Toom43 exists too! I announced it on this list some months ago :-) but the
  list was silent that times:
  http://gmplib.org/list-archives/gmp-devel/2009-February/000818.html .
  
Oh, it completely slipped between the cracks.  (I was extremly busy in
that period, with teaching and own studies.)

Is it OK if I invoke your FSF assign-future and put this into the GMP
repository, and release it in GMP 4.4?

  >   Marco Bodrato did produce a corresponding graph some time ago.
  >
  > Where?
  
  It's a .dvi document, not very elegant, but it can be evocative... It is
  on-line on my web page: http://bodrato.it/software/toom.html#TCcomp

Cute!  Even old toom62 has a place in this world, I had condemned it as
always slow.

It is interesting to see that the currently non-deplayed toom functions,
62, 52, 53, and 43 cover a very large part of the diagram's area.  Room
for improvements!

Diagrams are wonderful, they often bring more insight than more
theoretical analysis!

Thanks, Marco!

-- 
Torbjörn