PARI/GP and parallelism

PARI/GP and parallelism

[Andreas Enge]

Hello,

PARI/GP ("guix edit pari-gp") provides several mechanisms of parallelism,
which could be honoured by providing different package variants. 

After discussion with the author, I would suggest to enable the pthread
interface in the standard package (and to not provide an MPI variant unless
someone voices demand). This can be obtained trivially be adding the
configure flag "--mp=pthread". It will transparently parallelise a number
of internal functions according to the number of processor cores, and
should have negligible impact on single-core systems (which hardly exist
nowadays anyway).

What do you think? Are there any objections?

Andreas

Re: PARI/GP and parallelism

[zimoun]

Hi Andreas,

On Fri, 17 Jun 2022 at 14:35, Andreas Enge <andreas@enge.fr> wrote:

> configure flag "--mp=pthread". It will transparently parallelise a number
                     ^
                  --mt=pthread


> What do you think? Are there any objections?

No objection.  Maybe it is the occasion to also transform the previous
style with the new style for inputs [1].

Well, I do not know if the parallel computation of PARI/GP are
reproducible.  But the building of PARI/GP is not; because some
documentation…

--8<---------------cut here---------------start------------->8---
$ diff -r --no-dereference /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4{,-check}
Binary files /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4/share/pari/doc/develop.dvi and /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4-check/share/pari/doc/develop.dvi differ
Binary files /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4/share/pari/doc/INSTALL.dvi and /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4-check/share/pari/doc/INSTALL.dvi differ
Binary files /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4/share/pari/doc/libpari.dvi and /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4-check/share/pari/doc/libpari.dvi differ
Binary files /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4/share/pari/doc/parallel.dvi and /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4-check/share/pari/doc/parallel.dvi differ
Binary files /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4/share/pari/doc/refcard.dvi and /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4-check/share/pari/doc/refcard.dvi differ
Binary files /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4/share/pari/doc/refcard-ell.dvi and /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4-check/share/pari/doc/refcard-ell.dvi differ
Binary files /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4/share/pari/doc/refcard-lfun.dvi and /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4-check/share/pari/doc/refcard-lfun.dvi differ
Binary files /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4/share/pari/doc/refcard-mf.dvi and /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4-check/share/pari/doc/refcard-mf.dvi differ
Binary files /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4/share/pari/doc/refcard-nf.dvi and /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4-check/share/pari/doc/refcard-nf.dvi differ
Binary files /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4/share/pari/doc/tutorial.dvi and /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4-check/share/pari/doc/tutorial.dvi differ
Binary files /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4/share/pari/doc/tutorial-mf.dvi and /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4-check/share/pari/doc/tutorial-mf.dvi differ
Binary files /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4/share/pari/doc/users.dvi and /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4-check/share/pari/doc/users.dvi differ
--8<---------------cut here---------------end--------------->8---

…which seems a timestamp, as diffoscope says,

--8<---------------cut here---------------start------------->8---
--- /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4/share/pari/doc/INSTALL.dvi
+++ /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4-check/share/pari/doc/INSTALL.dvi
@@ -1,10 +1,10 @@
 00000000: f702 0183 92c0 1c3b 0000 0000 0447 1b20  .......;.....G. 
 00000010: 5465 5820 6f75 7470 7574 2032 3032 322e  TeX output 2022.
-00000020: 3036 2e31 373a 3133 3538 8b00 0000 0100  06.17:1358......
+00000020: 3036 2e31 373a 3134 3233 8b00 0000 0100  06.17:1423......
 00000030: 0000 0000 0000 0000 0000 0000 0000 0000  ................
 00000040: 0000 0000 0000 0000 0000 0000 0000 0000  ................
 00000050: 0000 00ff ffff ff8d 9ff2 0000 8ea0 024b  ...............K
 00000060: 661b 8da0 fdbe 99e5 8d92 00ae f267 f332  f............g.2
 00000070: 58ab 510b 000e 6666 000c 0000 0005 636d  X.Q...ff......cm
 00000080: 7231 32dd 4170 7090 6443 656e 6469 7891  r12.App.dCendix.
 00000090: 04b3 2f41 3a8e 9f15 477f 8d91 54b5 7449  ../A:...G...T.tI
--8<---------------cut here---------------end--------------->8---

Maybe it could be worth to tackle these. :-)


1: <https://guix.gnu.org/en/blog/2021/the-big-change/>


Cheers,
simon

Re: PARI/GP and parallelism

[Andreas Enge]

Am Fri, Jun 17, 2022 at 04:36:54PM +0200 schrieb zimoun:
> On Fri, 17 Jun 2022 at 14:35, Andreas Enge <andreas@enge.fr> wrote:
> > configure flag "--mp=pthread". It will transparently parallelise a number
>                   --mt=pthread

Indeed, thanks!

> No objection.  Maybe it is the occasion to also transform the previous
> style with the new style for inputs [1].

Yes, good point.

> Well, I do not know if the parallel computation of PARI/GP are
> reproducible.  But the building of PARI/GP is not; because some
> documentation…
> @@ -1,10 +1,10 @@
>  00000000: f702 0183 92c0 1c3b 0000 0000 0447 1b20  .......;.....G. 
>  00000010: 5465 5820 6f75 7470 7574 2032 3032 322e  TeX output 2022.
> -00000020: 3036 2e31 373a 3133 3538 8b00 0000 0100  06.17:1358......
> +00000020: 3036 2e31 373a 3134 3233 8b00 0000 0100  06.17:1423......

Time stamps in texlive. This is another kind of problem...

Andreas

Re: PARI/GP and parallelism

[Andreas Enge]

Am Fri, Jun 17, 2022 at 04:44:45PM +0200 schrieb Andreas Enge:
> Am Fri, Jun 17, 2022 at 04:36:54PM +0200 schrieb zimoun:
> > On Fri, 17 Jun 2022 at 14:35, Andreas Enge <andreas@enge.fr> wrote:
> > > configure flag "--mp=pthread". It will transparently parallelise a number
> >                   --mt=pthread
> Indeed, thanks!

Done.

> > No objection.  Maybe it is the occasion to also transform the previous
> > style with the new style for inputs [1].
> Yes, good point.

Someone else did the ordinary inputs, and the texlive native input looks
complicated, so I did not touch it.

Andreas

Re: PARI/GP and parallelism

[Maxime Devos]

[-- Attachment #1: Type: text/plain, Size: 672 bytes --]

zimoun schreef op vr 17-06-2022 om 16:36 [+0200]:
> Well, I do not know if the parallel computation of PARI/GP are
> reproducible.  But the building of PARI/GP is not; because some
> documentation…
> 
> --8<---------------cut here---------------start------------->8---
> $ diff -r --no-dereference /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4{,-check}
> Binary files /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4/share/pari/doc/develop.dvi and /gnu/store/fw2mckl5bn5mb6z79a1fyficvsclzdv3-pari-gp-2.13.4-check/share/pari/doc/develop.dvi differ

Maybe solved by
<https://lists.gnu.org/archive/html/guix-devel/2022-06/msg00171.html>!

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 260 bytes --]