P-1 on small exponents

[storm5510]

I have an old HP with a 2.8 GHz P4. It's been doing trial factoring. After looking at the PrimeNet summary and reading here, I plan to change it to something else.

In my account page on mersenne.org, I have seen a CPU option for P-1 small, but it's marked as "future". Based on what I see above, this can be set up manually.

It's sort of the same situation for my small laptop. It has an Intel Atom CPU and has also been doing trial factoring.

I've seen something referred to as "ECM Small". Not sure what that is.

My primary computer, Core 2 Duo, is running P-1 now with lots of RAM allocated. I would like to get away from factoring on the other two, if possible.

[Mini-Geek]

Originally Posted by storm5510

I've seen something referred to as "ECM Small". Not sure what that is.

Basically, it finds factors of small Mersenne numbers that are already known to be composite (and may or may not already have known factors).
ECM is one of the major steps in finding full factorizations of numbers. It is similar to P-1 in that it has a low-memory step 1 and a high-memory step 2, but each run of ECM has a low chance, and you can run it many times with the same B1 and B2 values with little wasted. (unlike P-1 in which multiple runs at the same B1 and B2 values is useless)

[markr]

Originally Posted by storm5510

I have an old HP with a 2.8 GHz P4. It's been doing trial factoring. After looking at the PrimeNet summary and reading here, I plan to change it to something else.

In my account page on mersenne.org, I have seen a CPU option for P-1 small, but it's marked as "future". Based on what I see above, this can be set up manually.

Yes, it's possible manually (but still using PrimeNet so they aren't grabbed by ECM workers) with some messing about. I can put together some detail of how I did it if anyone wants. On exponents with "inadequate" P-1, it definitely seems more productive than ECM-S, for finding factors of small Mersenne numbers. With exponents from 1.4M to 2.4M about 1 in 30 tests found a factor. (See the first post for bounds.)

It's sort of the same situation for my small laptop. It has an Intel Atom CPU and has also been doing trial factoring.

TF seems to be the logical choice, but if you're happy with the throughput at something else then it's up to you.

My primary computer, Core 2 Duo, is running P-1 now with lots of RAM allocated. I would like to get away from factoring on the other two, if possible.

Good idea - GIMPS has more TF power than it needs at the moment. If this PC has the RAM for PM1-L you could do that, or double-checks or first-time tests - it's up to you.

The 1.7GHz P4 that did most of my PM1-S is now doing a double-check, then it will do more PM1-L. The big benefit of P-1 (large) is that not only does it save a first-time-LL machine from doing the P-1, it also saves it from doing the last TF level by letting that work go to the well-resourced TF queue where it belongs.

[storm5510]

Originally Posted by markr

The 1.7GHz P4 that did most of my PM1-S is now doing a double-check, then it will do more PM1-L. The big benefit of P-1 (large) is that not only does it save a first-time-LL machine from doing the P-1, it also saves it from doing the last TF level by letting that work go to the well-resourced TF queue where it belongs.

My old HP has 1GB of RAM. That's the maximum for it. The CPU is 2.8 GHz. It has XP Pro on it, and a few device drivers. No other software. It doesn't have a lot of other processes running in the background.

[markr]

512MB is plenty for WinXP if no program needs a large amount of memory & you don't have 20 open at the same time. A lot of old computers with XP came with only 256MB, which was kind of enough for basic usage. So I reckon you could give prime95 enough for P-1. Remember to allow for memory used by integrated graphics, if relevant.

[storm5510]

It's got Intel integrated graphics. Its onboard is only 8 MB, but it will borrow another 56 MB from the main RAM to get to 64 MB. With only one core, I gave P95 half of the total RAM. 512 MB.

Is there any advantages of using a very large memory allocation? My primary computer is using 326 MB on Core 0 and 302 MB on Core 1.

It's at 88% on its last TF job. I just happened to stumble across an item under the "Test" menu labeled "Worker Windows". I thought it had something to do with the interface until I opened it. That's how I changed my work type instead of going to my GIMPS account.

I don't think that caption is very descriptive of what it actually does. Perhaps a suggestion to change it is in order.

[markr]

Originally Posted by storm5510

Is there any advantages of using a very large memory allocation? My primary computer is using 326 MB on Core 0 and 302 MB on Core 1.

There's discussion in this thread about P-1, memory and other things. Really really roughly, with more memory it goes a bit faster, but the bounds-choosing algorithm factors that in and the main effect is a better chance of finding a factor. Over 500MB probably yields very small benefits. That's per worker.

[storm5510]

I won't change the settings. The memory usage on the Core2 hasn't changed. The HP started on its P-1 about six hours ago. Stage 1. The CPU is running a little warmer, but not an issue.