Re: gp2c question

[John Cremona <john.cremona@gmail.com>]

On Fri, 11 Sep 2020 at 09:19, Bill Allombert
<Bill.Allombert@math.u-bordeaux.fr> wrote:
>
> On Fri, Sep 11, 2020 at 08:42:48AM +0100, John Cremona wrote:
> > Thanks Bill.  As a follow-up question, is it possible to use the
> > parallel computation system with gp2c?
>
> Of course! In fact it is sometime a bit easier than with GP.
> However parfor is only available for GP2C in the master PARI branch.

Of course -- I am using 2.12.0 (alpha) compiled yesterday but I forgot
to compile with the necessary flags so will rebuild.

>
> > I have a very simple script
> > which is doing a Monte Carlo sampling, and obviously I want to sample
> > as many points as I can.  The script takes 28m7s  for 10^9 samples and
> > 10x that (actually 4h27m20s) for 10^10.  If I could run it in parallel
> > x10 then I could get to 10^11 without waiting too long.  Of course I
> > can also run the script I have 10 times in parallel using gnu parallel
> > or similar -- perhaps that is easier anyway.
>
> Try to use parapply in GP, and then compile with GP2C.
> See
> https://pari.math.u-bordeaux.fr/Events/PARI2019/talks/parallel.pdf

I remember your talk!  But I have not had a real reason to use it yet.
Thanks again,

John

>
> Cheers,
> Bill.
>