Fwd: Mlucas + GMP

[Win C]

I have tested this FFT program for doing large multiplications, and it turns out that it is much faster starting from numbers like 2^400000. Should we consider integrating it into the GMP? That would definitely give us a boost. :D


-------- Original message --------
From: Win C <winsto003 at hotmail.com>
Date: 07/02/2018 14:48 (GMT+08:00)
To: "E.Mayer" <ewmayer at aol.com>
Subject: Re: Mlucas + GMP

Hi Ernst,

I would say the GMP developers would definitely be avoiding the GWNUM due to its restrictions on usage ( especially for finding primes ). Therefore, your mlucas would definitely be better.

I would email to the mailing list of GMP to see if they are interested. Thanks!


-------- Original message --------
From: "E.Mayer" <ewmayer at aol.com>
Date: 07/02/2018 14:45 (GMT+08:00)
To: Win C <winsto003 at hotmail.com>
Subject: Re: Mlucas + GMP

Hi, Winston -

I'm afraid I simply don't have the bandwidth for such an integration, and what gets added to GMP is out of my hands to begin with.  Were the GMP decision-makers (Granlund, Mo"ller, Bodrato, Zimmerman etc) interested in such a thing and willing to do most of the code-adaptation work themselves I would consider helping with it, but given that none of them has ever approached me about such a potential effort, I doubt they're interested.

But feel free to ask them whether they've ever considered integrating a non-GMP FFT freeware library - besides mine there is also George Woltman's GWNUM port of his Prime95 FFT routines, which is super-fast but more restricted in being x86-only.

Best regards,
-Ernst

On Feb 6, 2018, at 6:32 AM, Win C wrote:

Hi Ernst!

I am Wetitpig, one who is interested in the GNU MP Library. Could you please think of integrating your mlucas FFT code into the GMP library? That would definitely give the GMP library a speed boost. Thanks!

Or else, can you please describe in detail what is done within mlucas? Thanks!