speed of mpn_sqrtrem vs mpn_rootrem

Marco Bodrato bodrato at mail.dm.unipi.it
Thu Jul 30 15:21:40 UTC 2015


Ciao Paul,

Il Mer, 22 Gennaio 2014 1:00 pm, Zimmermann Paul ha scritto:
> tarte% ./speed -s 1000,3000,10000 -r mpn_sqrtrem mpn_rootrem.2
> overhead 0.000000002 secs, precision 10000 units of 3.53e-10 secs,
> CPU freq 2833.00 MHz
>           mpn_sqrtrem mpn_rootrem.2
> 1000      0.000153504       #0.9454
> 3000      0.000816813       #0.9187
> 10000     0.004606970       #0.9623

There are some more experiment that we can try, but the anomaly you
spotted should be healed now:
$ tune/speed -s 1000,3000,10000 -r mpn_sqrt mpn_root.2 mpn_sqrtrem
mpn_rootrem.2
overhead 0.000000002 secs, precision 10000 units of 2.86e-10 secs, CPU
freq 3500.08 MHz
             mpn_sqrt    mpn_root.2   mpn_sqrtrem mpn_rootrem.2
1000     #0.000051516        1.3866        1.4643        2.1827
3000     #0.000295437        1.3168        1.3855        1.8478
10000    #0.001850352        1.2448        1.3149        1.7012

You are welcome to test and/or give further suggestions

Best regards,
m

-- 
http://bodrato.it/



More information about the gmp-devel mailing list