Fast Multiplication
David Baumgärtel
baumgaertel.david at googlemail.com
Tue Nov 24 08:08:04 UTC 2015
Thank you very much!
The file "speed.c" in /tune was exactly what I needed.
Now I have a question regarding the size parameter -s.
I tried to find out how s is measured. Am I right that s is defined as 64
on a 64-bit CPU? And how is a number looking like with for example s=1 or
s=10?
Greetings,
David
Am 20.11.2015 09:17 schrieb "Torbjörn Granlund" <tg at gmplib.org>:
> David Baumgärtel <baumgaertel.david at googlemail.com> writes:
>
> I am a german student and I have a problem regarding the multiplication
> algorithms. I need to adjust the thresholds "MUL_KARATSUBA_THRESHOLD" and
> so on to force the usage of a particular multiplication algorithm. But
> that
> doesn't work like I want it to work, because the .h and .c files in which
> the algorithms are defined are precompiled.
> Is there any chance of changing the thresholds that I can compare the
> run-time of the different multiplication algorithms for equal inputs.
> Thank you very much for your effort!
>
> If you see MUL_KARATSUBA_THRESHOLD, then you have ages old source code.
> Please start by upgradig to a contemporary version of GMP.
>
> I am not sure what you mean by that .h and .c files are "precompiled".
> You should get GMP sources and compile the library yourself.
>
> For timing experiments, compile tume/speed. It's not docuimented, but
> the sources should be readable enough. This program can directly
> compare the speed of the various underlying multiplicaiton primitives.
>
>
> --
> Torbjörn
> Please encrypt, key id 0xC8601622
>
More information about the gmp-discuss
mailing list