basic benchmarks

[Karim Belabas <Karim.Belabas@math.u-psud.fr>]

Hi pari-dev,

  I have hacked together a quick bench for the arithmetic (integer and
floating point) in the native and GMP kernels. Results are available
for both 32bit and 64bit architectures from

  http://www.math.u-psud.fr/~belabas/pari/#bench

Or, specifically

  http://www.math.u-psud.fr/~belabas/pari/benchs/timings-int.html
  http://www.math.u-psud.fr/~belabas/pari/benchs/timings-mpfr.html

The native kernel doesn't fare too well compared to the GMP one. There are
compatibility issues with libpari, but I don't see any reason why gp
shouldn't use the GMP kernel by default nowadays. (Also, the current
Configure logic makes it cumbersome to disable gmp for libpari and
enable it for gp...).

Cheers,

    Karim.

P.S: as for integer arithmetic, pari-2.1.6 is about as fast as pari-2.2.10 +
native kernel [ except bezout() which is much slower ], so I did not
include it. The benches start at 100 decimal digits: all timings are
essentially identical at 10-20 digits, whatever the version and kernel.
-- 
Karim Belabas                     Tel: (+33) (0)1 69 15 57 48
Dep. de Mathematiques, Bat. 425   Fax: (+33) (0)1 69 15 60 19
Universite Paris-Sud              http://www.math.u-psud.fr/~belabas/
F-91405 Orsay (France)            http://pari.math.u-bordeaux.fr/  [PARI/GP]