Prime95 version 25.4
 

The Windows and Linux versions of 25.4 is now available. If you are running version 25.3 you can install the new version on top of the old version.

I need a few folks to do load testing on the new server. If you are interested, email or PM me. I'll PM those who have expressed interest with instructions. Load testers will be doing trial factoring of really big Mersennes to low limits or ECM on small Mersennes.

Steps to install version 25.4
----------------------------

1) DO NOT INSTALL OVER VERSION 24. If you have problems, you want to be able to revert to testing using version 24. There will be outages on the new server so you'll want to let version 24 run at lower priority to soak up those available CPU cyles.
2) Create a new directory, download [url]ftp://mersenne.org/gimps/p95v254.zip[/url], unzip into this new directory.
3) PM or email to get the secret line to add to prime.txt so that this version contacts the new server.
4) Run version 25.4. Follow the link in the dialog box to create a userid. Choose a work preference of either "Trial factoring to low limits" or "ECM on small Mersenne numbers".
5) Find problems. Report them to me.

Links to other versions
---------------------
Linux - [url]ftp://mersenne.org/gimps/mprime254a.tar.gz[/url]
Linux 64-bit - [url]ftp://mersenne.org/gimps/mprime254a-linux64.tar.gz[/url]
Windows 64-bit - [url]ftp://mersenne.org/gimps/p64v254.zip[/url]
Mac OS X - [url]ftp://mersenne.org/gimps/mprime254-MacOSX.tar.gz[/url]

Say we have all 4 windows open, in cascade mode.

Is there a button sequence that will let us page through the windows with the keyboard?

Even "F1", "F2", F3" and "F4" would be cool, or a single character toggle.

If we tile the windows they are too small on our screen.

[SIZE=1](Ideally, we'd like plain old tabs, like Firefox, but we understand why UI design is not that important, and that you cannot make everyone happy.)
[/SIZE]
:smile:

[QUOTE=Xyzzy;112256]Is there a button sequence that will let us page through the windows with the keyboard?[/QUOTE]

It looks like CTRL+F6 does this

I'm running a quad-core machine, and I'm wondering if there are plans to be able choose a preference for multiple types of work on the same machine. For instance, it would be cool if I could do P-1 factoring on one core, ECM on another, run LL tests on the third, and keep the 4th core open for some other project. I don't know how popular this would be, but I think it's interesting to being doing different types of work at once.

Also, one reason you might want to have the ability to stop and start different threads independently is if you're going to do something memory-intensive for a short period of time and want to suspend the thread that is P-1 factoring but let a trial factoring or LL thread continue.

Here's a more speculative wish list item, so let me know if there's some reason this wouldn't be as useful as I'm thinking. It looks like Primenet v5 is going to become more of a center for finding first factors of Mersennes in addition to the main GIMPS project of discovering Mersenne Primes. P-1 factoring is special in that if you want to extend to higher limits, you have to redo all of the work that someone else did in P-1 factoring to a lower limit. If you had their save file, you could save a little bit of time (perhaps not enough to be worth it). This might be most useful for smaller Mersennes (<100k?) where a good bit of time is spent in getting to the current P-1 limits. I'm not sure how serious the bandwidth and storage requirements would be for transferring these save files, so it might not be worth it for a secondary project.

Also, I'm not sure exactly where the factoring crossovers are here. Are we at the point where we should forget about P-1 for Mersennes less than (say) 50k and just do ECM? It seems like P-1 is still useful for Mersennes closer to 500k and higher, but i'm curious where the crossover is. Thanks.

[QUOTE=jbristow;112259]I'm wondering if there are plans to be able choose a preference for multiple types of work on the same machine. For instance, it would be cool if I could do P-1 factoring on one core, ECM on another, run LL tests on the third, and keep the 4th core open for some other project.[/QUOTE]

You can do that in the Test/Worker Threads dialog box. The Worker number drop down box lets you select which worker window you'd like to set work preference for.

[quote]It looks like CTRL+F6 does this[/quote]

:love:

[QUOTE=Prime95;112154]Follow the link in the dialog box to create a userid. Choose a work preference of either "Trial factoring to low limits" or "ECM on small Mersenne numbers".[/QUOTE]
I'm trying this, but I don't see these options. I'm seeing the usual options:

Whatever makes the most sense
10,000,000 digit numbers to test
First time tests
Double-check tests
Trial factoring

Edit: never mind I got it working. I had create and add that line to prime.txt before prime.txt is created for it to use the new server.

[QUOTE=jbristow;112259]It looks like Primenet v5 is going to become more of a center for finding first factors of Mersennes in addition to the main GIMPS project of discovering Mersenne Primes. P-1 factoring is special in that if you want to extend to higher limits, you have to redo all of the work that someone else did in P-1 factoring to a lower limit.[/QUOTE]

A P-1 run is worth the same amount as an ECM curve to the same limits -- except that P-1 for Mersennes gets the exponent for "free" as the exponent will divide any factor-1. If you don't understand that, suffice it to say that it makes a P-1 to B1 worth the same as an ECM run to a B1 about 5 to 10 times larger.

What that all means is that I'm not prepared to hassle with saving P-1 intermediate files. It's easier to just assign a few ECM curves to get the equivalent amount of factor finding power.

[quote=Prime95;112270]A P-1 run is worth the same amount as an ECM curve to the same limits -- except that P-1 for Mersennes gets the exponent for "free" as the exponent will divide any factor-1. If you don't understand that, suffice it to say that it makes a P-1 to B1 worth the same as an ECM run to a B1 about 5 to 10 times larger.

What that all means is that I'm not prepared to hassle with saving P-1 intermediate files. It's easier to just assign a few ECM curves to get the equivalent amount of factor finding power.[/quote]

Okay, that makes sense. P-1 gets a bump from the special 2kp+1 format of Mersenne Factors, but in the end the advantages of being able to do multiple ECM runs at the same B1 and its ability to find unsmooth factors makes it not worth over-optimizing P-1 searches to large B1s.

The "Assignment details" page on the new server sometimes gives an incomplete listing.

I've got two cores, one each assigned to ECM and TF. The client always shows 15 assignments of each type, but the Primenet server often cuts the list of TF tests short.

Two more things...

The "Worker threads" dialog doesn't update the "worker-specific options" behavior when you change the number of threads. For instance, when I change from 1 thread to 2, the "worker number" field is still greyed-out. You have to close the dialog and open it again before you can change those options.


Also, I just tried the new software with my hyperthreaded P4 machine. It defaulted to 1 thread, using 2 'CPUs' per thread. When I tried to change this to run 2 threads using 1 'CPU' apiece, it would not fetch any work from the v5 server for the second thread. So I can only run one test at a time. Is this intended behavior?