Prime95 or GMP-ECM
 

I had a few questions regarding the performance of the different implementations of ECM. I apologize if these questions have been answered here before but a quick check on the forum did not reveal any answers.

1. For running ECM on small Mersenne numbers, is Prime95 better than GMP-ECM?

2. Are the pre-compiled Linux 64-bit binaries of GMP-ECM the fastest for a general i.e. non-mersenne number? Basically, is Linux better than Windows? And is 64-bit better than 32-bit?

3. Does GMP-ECM speak to all the ECMNet servers (given the IP address and port) that have been posted on this forum? Such as xilman's server?

Thanks.

[QUOTE=garo;231912]I had a few questions regarding the performance of the different implementations of ECM. I apologize if these questions have been answered here before but a quick check on the forum did not reveal any answers.

1. For running ECM on small Mersenne numbers, is Prime95 better than GMP-ECM?

2. Are the pre-compiled Linux 64-bit binaries of GMP-ECM the fastest for a general i.e. non-mersenne number? Basically, is Linux better than Windows? And is 64-bit better than 32-bit?

3. Does GMP-ECM speak to all the ECMNet servers (given the IP address and port) that have been posted on this forum? Such as xilman's server?[/QUOTE]

1. IIRC, the latest GMP-ECM can use gwnum in addition to GMP.

2. 64-bit is better than 32-bit, hands down. Linux/Windows shouldn't make a huge difference outside of OS overhead.

3. GMP-ECM does not communicate with ECMNet. The ECMNet client can use any version of GMP-ECM. I don't know which version of ECMNet Paul is using, but I presume it is one of my releases (2.x) as I added support for P+1/P-1, which the old ECMNet did not support.

[QUOTE=rogue;231915]
3. GMP-ECM does not communicate with ECMNet. The ECMNet client can use any version of GMP-ECM. I don't know which version of ECMNet Paul is using, but I presume it is one of my releases (2.x) as I added support for P+1/P-1, which the old ECMNet did not support.[/QUOTE]To be precise, ECMnet is a client and a server. The client calls a program to do the factoring work. That client could, in principle, be any program at all as long as it produces results which the client can understand. In practice, GMP-ECM is the program which the client uses.

I can't remember which version of the server I'm running. It's certainly a fairly recent 2.x server.


Paul

Thanks all. But back to Q1. Is gwnum faster for both stages on a small Mersenne number, or only stage 1 or none?

[QUOTE=rogue;231915]1. IIRC, the latest GMP-ECM can use gwnum in addition to GMP.[/QUOTE]
Haven't noticed any compiled binaries. hint hint:smile:

[QUOTE=xilman;231920]
I can't remember which version of the server I'm running. It's certainly a fairly recent 2.x server.
[/QUOTE]

I have just noticed that the [URL="http://www.loria.fr/~zimmerma/records/ecm/implementations.html"]ECMNet-homepage[/URL] say that Mark Rodenkirch has released version 2.6.1 of the server/client utility, but the link given there is 404.

[QUOTE=garo;231923]Thanks all. But back to Q1. Is gwnum faster for both stages on a small Mersenne number, or only stage 1 or none?[/QUOTE]

gwnum is faster in stage 1 only.

There is a way to configure prime95 to just do stage 1 and then run GMP-ECM to do stage 2.

[QUOTE=Andi47;231929]I have just noticed that the [URL="http://www.loria.fr/~zimmerma/records/ecm/implementations.html"]ECMNet-homepage[/URL] say that Mark Rodenkirch has released version 2.6.1 of the server/client utility, but the link given there is 404.[/QUOTE]

Try this link: [url]http://home.roadrunner.com/~mrodenkirch/home/ECMNet.html[/url]
The latest version is apparently 2.7.3.

Speaking of which, I've made some modifications to the client which I have found to be very useful, particularly when wanting to run multiple client processes which share a config file. Mark, are you accepting changes of this sort?

[QUOTE=Prime95;231930]gwnum is faster in stage 1 only.

There is a way to configure prime95 to just do stage 1 and then run GMP-ECM to do stage 2.[/QUOTE]

Thanks George. I noticed that those lines in undoc.txt have not been modified so I suspected this might be the case. Thanks to everyone else for responding.

[QUOTE=jyb;231933]Try this link: [url]http://home.roadrunner.com/~mrodenkirch/home/ECMNet.html[/url]
The latest version is apparently 2.7.3.

Speaking of which, I've made some modifications to the client which I have found to be very useful, particularly when wanting to run multiple client processes which share a config file. Mark, are you accepting changes of this sort?[/QUOTE]

Possibly. I've started a [URL="http://www.mersenneforum.org/showthread.php?t=13996"]new thread[/URL].

Someday I'll have to notify Paul Zimmerman of the current version so that he can update his link.

[QUOTE=garo;231946]Thanks George. I noticed that those lines in undoc.txt have not been modified so I suspected this might be the case. Thanks to everyone else for responding.[/QUOTE]

You add "GmpEcmHook=1" to prime.txt, and then set B2 value to 1.

Then you use gmp-ecm on results.txt like this with B1=1 and B2 to whatever you want:

ecm -resume results.txt 1 <B2>