Karim BELABAS on Thu, 12 Dec 2002 23:24:04 +0100 (MET)


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

Re: 2 changes in the CVS


On Fri, 6 Dec 2002, Igor Schein wrote:
> On Fri, Dec 06, 2002 at 03:12:55PM +0100, Karim BELABAS wrote:
>> On Fri, 6 Dec 2002, Bill Allombert wrote:
>>> I have commited two change to the CVS.
>>
>> While we are at it: I have enabled Karatsuba multiplication for t_REAL (code
>> has been here for at least 2 years, but was unused). It is not properly tuned
>> yet, but at least it seems to give correct (and correctly rounded) results.
>
> My testsuite runs about 10% slower now.  I'll wait 'til it's tuned.
> Or will a slowdown for default precision unavoidable?

I've included KARATSUBA_MULR_LIMIT to the list of defines computed by the
'tune' utility [ make tune && tune ]. It turns out that full product using
Karatsuba mulii() becomes consistently faster than default mulrr() for
arguments having about 300 words (!) on my machine.
[ should really implement Mulders/Hanrot-Zimmermann short products... ]

Back to normalcy ?

    Karim.
-- 
Karim Belabas                    Tel: (+33) (0)1 69 15 57 48
Dép. de Mathématiques, Bât. 425  Fax: (+33) (0)1 69 15 60 19
Université Paris-Sud             Email: Karim.Belabas@math.u-psud.fr
F-91405 Orsay (France)           http://www.math.u-psud.fr/~belabas/
--
PARI/GP Home Page: http://www.parigp-home.de/