Philippe de Rochambeau on Fri, 19 Nov 2021 08:10:38 +0100


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: GP runs on iOS


Hello,

I carried out a few benchmarks, this morning.

1) factor(2^256-1)

On an iPad Air 3, A12 Bionic Chip running a 32-bit i86 Alpine Linux Virtual Machine, GP 2.14.0 single thread with parisize 4000000, primelimit = 500000, factorization took 2:06.77 minutes.

On a MacBook Pro 2,6 GHz Intel Core i7 6 core, GP 2.13.3, 12-thread, with arisize = 8000000, primelimit = 500000, factor(2^256+1) factorization was instantaneous.

On an Android Galaxy 6+, PariDroid 2.13.2.0 took 14 seconds to carry out the computation.

On Windows 10, Dell Core i7-10510U, 2.13.3 single-threaded GP took about 2 seconds to factorize.


2) Caesar

alphabet = Vecsmall("ABCDEFGHIJKLMNOPQRSTUVWXYZ")
get(v,i) = v[((i)%#v)+1]
caesar(s,n) = {
    n1 = [get(n,i + s) | i <- [0..#n - 1]];
    strchr(n1)
}
c = caesar(3, alphabet)
print(c)


… was instantaneous on both the iPad and the MacBook Pro

On Android, it caused ParidDroid to freeze

On Windows, it was instantaneous.

Best regards,

Philippe



Le 19 nov. 2021 à 04:13, Thomas Kellar <wangude@gmail.com> a écrit :

FWIW

GP/PARI CALCULATOR Version 2.11.2 (released)
running on Ubuntu Linux on a  Ryzen processor

factor(2^256+1)   takes about 1 second

thomas

On Thu, Nov 18, 2021 at 5:26 PM Bill Allombert
<Bill.Allombert@math.u-bordeaux.fr> wrote:

On Thu, Nov 18, 2021 at 10:55:23PM +0100, Philippe de Rochambeau wrote:
Hi Bill,
If you search for Pari on this website,
https://pkgs.alpinelinux.org/packages?name=pari&branch=edge
<https://pkgs.alpinelinux.org/packages?name=pari&branch=edge>, you’ll
notice that it is available for Alpine Linux, iSh’s underlying OS.
Unfortunately, a 32-bit version isn’t available, which is why I had to
compile the source code.

Well, at least in principle, this is easy to fix.

Right now, iSh won’t let you install another Linux distribution (Centos, etc.)

Please find enclosed make bench’s output.


factor(2^256+1) took 2:06.59 minutes.

This seems much slower than the Webassembly version.
<https://pari.math.u-bordeaux.fr/gpwasm.html>
though of course I did not try on your hardware, but it should be less
than 10s.

Cheers,
Bill.