GPGPU support

Joerg Arndt arndt at
Tue Oct 12 15:12:22 CEST 2010

* Morten Gulbrandsen <Morten.Gulbrandsen at rwth-Aachen.DE> [Oct 12. 2010 14:23]:
> [...]
> My question is: will many cores or paralell computing help with
> gmp-Chudnowski? Just for comparison? Why is not the number of cores
> reported? I assume all results was 64 bits as an increase from 32 bits
> to 64 bits will speedup with a factor four times.
> If so, (if multi-core speeds up) I think we should go for it, If Not,
> if the same results are achieved with one core as with many cores, I'll
> look for a different approach.

I'd suggest multi-core is the way to go.

Using the ideas given in
sect.21.10 (pp.438-439)
sect.22.5 (pp.451-454) [there especially sect.22.5.5]
of the fxtbook should lead to a multi-thread FFT which
can be made to work with external storage.
Note the float-FFT should be replaced by (e.g.)
3 NTTs (for the 3-primes algorithm).
Also look at sect.25.8 (pp.529-530):
the technique gave me a 50% speedup with
the multiplication for a high-precision
floating point library.

> Sincerely yours,
> 藻留天   具留部覧度船
> Morten Gulbrandsen
> [...]

More information about the gmp-discuss mailing list