What work does YOUR processor do?
 

How do you split your CPU resources for GIMPS? I made a template using the work types - AUTO (whatever makes the most sense), LL, D, P-1, TF, ECM, WRP (world record primes) so you can show everyone. I am currently running:

[CODE]Worker #0: LL
Worker #1: LL
Worker #2: LL
Worker #3: LL[/CODE]

I may very soon switch to a more 'useful' config of:

[CODE]Worker #0: LL
Worker #1: D
Worker #2: P-1
Worker #3: TF[/CODE]

As much as I want to find a mersenne, I can't help but feel that giving some of my other cores some factoring and double-checking work to do would be more useful for GIMPS.

I think there are several answers depending on your perspective...

1. The more straight forward answer might be to pick the option "Whatever makes the most sense" and let the server pick

2. Whatever intrigues you the most: some like the thrill of finding a prime; some like lots of quick results with factoring; some like ...

3. Whatever your PC is best suited for: i.e. a 100 Mhz PC would take VERY long to do a LL test but performs TF-LMH quite well. If you have less than 500 MB RAM free you probably don't want to do P-1. This site can help you choose: [url]http://mersenne-aries.sili.net/throughput.php[/url]

4. Whatever is requested help with: i.e. George has noted on another thread that he needs more workers on P-1 and DC and there are too many doing factoring. [url]http://www.mersenneforum.org/showpost.php?p=181129&postcount=239[/url]

5. If you are into stats and standings you could pick a less popular work type like ECM and move up quickly.

Personally over the years I've been in each of the 5 above camps at one time or another. Right now I'm mostly in 4 (DC and P-1).

DC/DC/DC/P-1

TF is overpowered so unless you have a good reason to run it, do P-1 on the last core instead.

[quote=hj47;207080]How do you split your CPU resources for GIMPS? I made a template using the work types - AUTO (whatever makes the most sense), LL, D, P-1, TF, ECM, WRP (world record primes) so you can show everyone. I am currently running:

[code]Worker #0: LL
Worker #1: LL
Worker #2: LL
Worker #3: LL[/code]I may very soon switch to a more 'useful' config of:

[code]Worker #0: LL
Worker #1: D
Worker #2: P-1
Worker #3: TF[/code]As much as I want to find a mersenne, I can't help but feel that giving some of my other cores some factoring and double-checking work to do would be more useful for GIMPS.[/quote]

On my Windows machine I'm running

Worker #0: LL
Worker #1: LL
Worker #2: LL
Worker #3: LL
on my Linux machine I'm running one 100M LL on 4 cores.

It maybe not the optimal setting but I want to win the price ^^ ...

Worker #0: PRP \
Worker #1: PRP > Five-or-Bust (it is now actually One-or-Bust)
Worker #2: PRP /
+
PFGW on CPU #3: PRP -> CRUS

consider the chances :smile::coffee:

not to nitpick, but shouldn't the template be:

[CODE]Worker #1:
Worker #2:
Worker #3:
Worker #4:[/CODE]

since this is how Prime95 displays them?

or if you want to be more precise:

[CODE]Core #0:
Core #1:
Core #2:
Core #3:[/CODE]

well for me since I am testing my overclock since I had to get a new motherboard (had two of the same model blow up, went to the next model up) I am running 4 double checks. But nominally I run
[CODE]Worker #1: P-1
Worker #2: WRP
Worker #3: WRP
Worker #4: D[/CODE]

Worker #0: LL
Worker #1: LL
Worker #2: LL
Worker #3: LL or P-1/TF

[QUOTE=hj47;207080]How do you split your CPU resources for GIMPS? I made a template using the work types - AUTO (whatever makes the most sense), LL, D, P-1, TF, ECM, WRP (world record primes) so you can show everyone.[/QUOTe] There are 2 GIMPS types that you left off TF-LMH and 100ML-L. I think that those that run the latter are likely to use multiple cores per test.

All of the machines that I normally handle are:
Worker#1: TF-100M (by hand)
Worker#2: TF-100M

The other machines that are borgim (and not always working) are:
Worker#1: TF
Worker#2: TF
(because they are borg and may or may not be running at any given moment, TF is better [I know it is over subscribed]).

I'm doing everything with my 10 cores.

I prefer ECM-F and P-1, the first for my research, the second per GIMPS needs.

Luigi

Unless you are really keen to find a prime, I would reccommend the setting proposed by garo on post #3: DC/DC/DC/P-1. These are the types of work GIMPS is in most need at the moment.
I am assuming your machine has an amount of memory suitable for P-1 work. If that is not the case, the fourth worker could either do DC, or, if you start noticing a slowdown due to memory access congestion, Trial Factoring.

On my slowest dual-core (AMD 4400) I run TF on one core, P-1 on the other.

Moving from slowest pc toward faster, on my next three dual cores (two PD-805s, and a PD-915) I run DC on one core, and P-1 on the other core.

The next two dual-cores (an E6400, and an E6700), I run first-time LL on both cores.

On the two quad-cores (Q9450, and an i7 920), I run first-time LL on three cores, and DC on the other core.

Out of 20 total cores - one TF, four P-1, five DC, and ten first-time LL. All are over-clocked from 10-30% . I have run multiple DCs on all of them to confirm reliability.

[quote=lycorn;207225]Unless you are really keen to find a prime, I would reccommend the setting proposed by garo on post #3: DC/DC/DC/P-1. These are the types of work GIMPS is in most need at the moment.
I am assuming your machine has an amount of memory suitable for P-1 work. If that is not the case, the fourth worker could either do DC, or, if you start noticing a slowdown due to memory access congestion, Trial Factoring.[/quote]

I don't see the 'fuzz' / problem with the P-1. If nobody else has done it before don't you just do it at the beginning of your LL test?

[quote=joblack;207257]I don't see the 'fuzz' / problem with the P-1. If nobody else has done it before don't you just do it at the beginning of your LL test?[/quote]
Yes, but how often will machines that are mainly used for LL allow enough memory to have a really worthwhile P-1 test? I don't really know, but I'm sure more factors are found when you get computers more dedicated towards P-1 to do that part.

True. Machines that are dedicated to LL testing are very often left with the default Prime95 memory allocation, which is substantially lower than the values allocated by somebody dedicating a given machine to P-1 work. Therefore, the probability of finding a factor is higher if you have "specialized" P-1 factor hunters.
Note as well that P-1 is performed on a given exponent before the final level of TF. If P-1 is done before the front wave of LL reaches the exponent, the TF will also be done by somebody working specifically on TF, so when the exponent is finally handed over to the LL tester, nothing else is left to do, and the tester can immediately proceed to the type of work he´s actually willing to do.
In the database one can find many examples of exponents LL tested, but with only Stage 1 performed, or even no P-1 test at all; these correspond to people allocating small amounts of memory to P-1 or even skipping the test altogether. Some of these tests (and the DC that will have to follow) could have been avoided had P-1 been properly done.
To summarize: the advantages of having P-1 as a separate assignment type are: a higher chance of finding factors, and to allow people to work just on what they really want, therefore raising their level of motivation.

mine are DC/P-1/LL/LL

2 machines - dc/dc/dc/p-1 and dc/p-1

I'd do more p-1 as I have plenty of ram - but the version of prime I'm running, mprime ignores the 'mem per thread' option int he config files. Not sure if this is fixed in later versions.

-- Craig

The current version (25.11) of Prime95 ignores it as well. George has written sometime ago that such behaviour will be corrected in the next version. But that shouldn´t be keeping you from running more than one P-1 test on a multi-core machine. The worker windows will share the available memory evenly, according to each other needs. I have been running, for several months now, P-1/P-1 on a dual core CPU with no problems at all. I´ve allocated 1200 Mb of mem to Prime95, and when one window is running Stage 1 and the other Stage 2, the latter takes around 1140 Mb. When both are running Stage 2, each one takes ~600 Mb. The program manages the mem allocation automatically

[QUOTE=lycorn;207847]The current version (25.11) of Prime95 ignores it as well. [/QUOTE]Website say it is version 25.9

25.11 is available here:
[URL]http://www.mersenneforum.org/showthread.php?t=12155[/URL]

I am new to PrimeNet and have recently enlisted a dedicated quad core machine with 4 GB of RAM to find a prime. Here're some basic questions:

1. When the cores first started working, it was looking for small factors. What is the acronym for this work?

2. I've seen the memory allocating activity on the screen so I assume this is P-1 work. Is this where the mersenne's equation is being calculated?

3. The results record show that P-1 is completed and now it is doing LL work.

4. What does DC, D, and ECM mean? I believe TF means trial factoring.

5. When a computer receives an exponent, does PrimeNet allow the same computer to continue working on the same number until it is determined to be a prime or not?

6. From the discussion above, it appears that a person who is strictly after the reward for finding a large prime number could just ask for LL assignments. Is this correct?

[QUOTE=esqrkim;209703]I am new to PrimeNet and have recently enlisted a dedicated quad core machine with 4 GB of RAM to find a prime. Here're some basic questions:

1. When the cores first started working, it was looking for small factors. What is the acronym for this work?
[/quote]
either TF or P-1
[quote]
2. I've seen the memory allocating activity on the screen so I assume this is P-1 work. Is this where the mersenne's equation is being calculated?
[/quote]
P-1 is another way to find relativly small factors.
[quote]
3. The results record show that P-1 is completed and now it is doing LL work.
[/quote]
this is not a question.
[quote]
4. What does DC, D, and ECM mean? I believe TF means trial factoring.
[/quote]
DC is Double Check. It basicly repeats the LL test to make sure no errors occured.
ECM stand for Eliptic Curve Method and is another way to find small factors.
[quote]
5. When a computer receives an exponent, does PrimeNet allow the same computer to continue working on the same number until it is determined to be a prime or not?
[/quote]
ya, so long as it keeps working on it a reporting in periodically it keeps that exponent exclusivly.
[quote]
6. From the discussion above, it appears that a person who is strictly after the reward for finding a large prime number could just ask for LL assignments. Is this correct?[/QUOTE]
Well, they could but that would be anti-social. The LL-only category is intended for people running other programs(other that prime95 or mprime, there are several) who do not have the TF or P-1 implemented to run exclusively LL tests. Also I guess if you have very little memory available you might prefer not to run P-1.

Edit: Oops, lfm posted several minutes before I finished composing this.

I have no quarrel with any of lfm's answers except the last one, in reply to esqrkim's #6. I don't see how esqrkim's #6 is anti-social. It's perfectly okay to want to test a number eligible for the next EFF prize. That's why the "LL-100M" type of request was made available in PrimeNet! Perhaps lfm may be confusing the "LL" assignment type with the "LL-NF" assignment type.

Or maybe I'm misinterpreting esqrkim's #6 and he _is_ asking about "LL-NF", and lfm's response _is_ appropriate. Only esqrkim can tell.

- - -

[quote=esqrkim;209703]1. When the cores first started working, it was looking for small factors. What is the acronym for this work?[/quote]The first factoring method used is trial factoring, TF. A second method, P-1 factoring (PM1), could also find "small" factors, depending on the definition of "small".

[quote]2. I've seen the memory allocating activity on the screen so I assume this is P-1 work. Is this where the mersenne's equation is being calculated?[/quote]No, but what you mean by "mersenne's equation" is unclear. There's not some standard "mersenne's equation".

P-1 is a factoring method invented by John Pollard. See [URL="http://en.wikipedia.org/wiki/Pollard%27s_p_%E2%88%92_1_algorithm"]http://en.wikipedia.org/wiki/Pollard%27s_p_%E2%88%92_1_algorithm[/URL] It uses a mathematical method called Fast Fourier Transform (FFT) to perform multiplications.

[quote]4. What does DC, D, and ECM mean?[/quote]DC (and D) = Double-checking. ECM = Elliptic Curve Method, another factoring method.

[quote]I believe TF means trial factoring.[/quote]Correct.

[quote]5. When a computer receives an exponent, does PrimeNet allow the same computer to continue working on the same number until it is determined to be a prime or not?[/quote]First of all, PrimeNet can't control what a user's computer does. You [I]can[/I] use any computer at any time to test a number for primality -- [U]if[/U] you don't care about overlapping/duplicating someone else's work, [I]which can cause hard feelings[/I] and constitutes "poaching" if done deliberately.

What PrimeNet [U]can[/U] do is assign work so that no one is "officially" working on the same number as anyone else who's working on a PrimeNet assignment. It's basically a voluntary cooperative system.

The contract between PrimeNet and you is:

1. In return for your requesting assignments and working only on assigned exponents, PrimeNet will do what it can to assure that it assigns no one else the same number at the same time.

2. You agree to report your result to PrimeNet in a timely manner so it can later on assign that number to someone else for further work, if needed.

3. If you don't report either a result, or progress toward a result, within the posted time limit (currently every 60 days), PrimeNet reserves the right to reassign the exponent, on the assumption that you've abandoned work on the assignment.

Now, there are different types of PrimeNet assignment. If PrimeNet assigns work as TF, it expects you to quit working on that number after you report your TF result (because PrimeNet will assume that after your TF report, it is free to assign that number to someone else for different work). Same for PM1 and ECM .

LL and DC(D) are a bit different -- if your prime95 program decides that not enough TF and PM1 work has been done yet on the exponent, it can proceed to finish up TF and/or PM1 before starting the long L-L (Lucas-Lehmer) test. This is okay with PrimeNet.

[quote]6. From the discussion above, it appears that a person who is strictly after the reward for finding a large prime number could just ask for LL assignments. Is this correct?[/quote]Almost.

Because the next unclaimed EFF prize is for numbers larger than 100 million digits, he needs to specify "LL-100M".

If he specifies "LL" or "LL-WR" or "LL-NF", he'll be assigned an exponent for a number smaller than 100 million digits. ("LL-10M" is now obsolete, meaning only the same as "LL", because [I]all[/I] current first-time assignments are exponents for more-than-10-million digit numbers.)

Since the current world record Mersenne prime has (many) less than 100 million digits, specifying "LL-WR" does _not_ guarantee being assigned an exponent eligible for the next EFF prize.

Okay. I can figure out LL-10M and LL-100M, but what is LL-NF?

My computer is set to recieve assignments automatically, that is, I haven't specified any preference. It appears that I am still working on the same 4 exponents and it is currently performing LL test.

How can I tell what kind of assignment was given to me?

[QUOTE=esqrkim;209753]Okay. I can figure out LL-10M and LL-100M, but what is LL-NF?

My computer is set to recieve assignments automatically, that is, I haven't specified any preference. It appears that I am still working on the same 4 exponents and it is currently performing LL test.

How can I tell what kind of assignment was given to me?[/QUOTE]

You have 2 ways to choose your kind of work:

1 - From Prime95 menu, Test/Worker windows.
2 - From PrimeNet, My Account/CPUs

Luigi

[quote=esqrkim;209753]How can I tell what kind of assignment was given to me?[/quote]Log in at [URL]http://mersenne.org/[/URL]

In your account summary, under heading "Workload", click on the "Assigned" link. The displayed table of your assignments shows the work type for each exponent.

[quote=esqrkim;209753]Okay. I can figure out LL-10M and LL-100M, but what is LL-NF?[/quote]Go back to the account summary page. Under "Workload", click on "Default account work-type".

Under "[B][FONT=Tahoma][SIZE=2]Select a different account-level Work Type preference here:",[/SIZE][/FONT][/B] click on the down-arrow to see the complete list of work types, including "LL-NF".

[quote]My computer is set to recieve assignments automatically, that is, I haven't specified any preference.[/quote]You can select a preference under "[B][FONT=Tahoma][SIZE=2]Select a different account-level Work Type preference here:",[/SIZE][/FONT][/B]

Thanks for the info.

A little off the topic, but I have a question about the info in my account. I have two computers listed in my account, but i am not using one of them. How can I update?

Does it matter? If one computer never reports results, then ... it won't have any results listed. If it never requests an assignment, it won't have assignments listed. If it erroneously has an assignment, you can release it.

But if you just want to clean up, look around the CPUs page -- there's something for deleting.

... I now have a serious case of computing envy.

Borged an old Geode NX 1500 to do TF.

When I'm around to keep it on a Sempron 2800+ to do TF (I'm not around it enough to give it anything better).

And Core 2 Duo that once it finishes a DC it will be going to P-1. NOTE: Yes it does only run one thread and it runs on a throttle of 50 or 33 due to being in a laptop and reaching disturbingly high temperatures (I've already melted one motherboard)

hey guys,

if i check my CPU on this site ([URL="http://mersenne-aries.sili.net/throughput.php?cpu=AMD+Athlon%28tm%29+64+X2+Dual+Core+Processor+4600%2B%7C512%7C0&mhz=2700"]http://mersenne-aries.sili.net/throughput.php?cpu=AMD+Athlon%28tm%29+64+X2+Dual+Core+Processor+4600%2B|512|0&mhz=2700[/URL])

it gives me an low efficiency for nearly all worktypes! I dont understand this :)

Hopfe you can help me :)

AMD Athlon 4600+ OC to 2700mhz

That's because your processor is much more efficient at factoring to low limits than all other worktypes when compared with the benchmark Core2. Don't let that stop you from doing other worktypes though as factoring to low limits is more or less finished.

Jeah ok, but why is my processor much more effivient at favtoring to low limits?!

Well, the most time i am doing DC :-D so no problem there ^^

[quote=Kai.s0r;210104]Jeah ok, but why is my processor much more effivient at favtoring to low limits?![/quote]Because of the differences between the way AMD designed the Athlon and the way Intel designed the Core2. Even when they execute the same instructions, their circuitry is different, and the two manufacturers made different tradeoffs in their designs.