Some mulrr timings.

[Bill Allombert <allomber@math.u-bordeaux.fr>]

Hello PARI testers,

I have made a small test of the recent mulrr improvments.

Here the test:
-------------
\p100
Pi;
for(i=1,10000,Pi*(Pi+1));
##
\p1000
Pi;
for(i=1,1000,Pi*(Pi+1));
##
\p10000
Pi;
for(i=1,100,Pi*(Pi+1));
##
\p50000
Pi;
for(i=1,10,Pi*(Pi+1));
##
---------------

Result:
             | p100     p1000    p10000     p50000
GP 2.1.4     : 140 ms.  250 ms.  2,160 ms.  5,580 ms.
GP 2.2.5     : 150 ms.  250 ms.  1,380 ms.  1,780 ms.
GP 2.2.5-GMP : 140 ms.  200 ms.    630 ms.    700 ms.

This show that the recent hack in the GMP kernel that allow mulrr
to use mpn_mul is really useful...

Cheers,
Bill.