Work for chosen exponent

[Petnek]

Hello,
is it possible to make calculations for chosen exponent? For example this one. It's the closest exponent to my date of birth and I would appreciate a chance to take care of it

Is here a way how to make calculations for it? If yes, how? And in which order? For example TF is not possible to reserve due to current active range.

Thank you

[LaurV]

That exponent generates a mersenne number that already has a factor, therefore there is nothing more to calculate for it. If you want to stay as close to it as possible, go to the link you posted, place your mouse over the little arrows in the upper-left table (where the exponent is printed in big letters) and look for "next exponent without factor" respectively "previous exponent without factor". And pick those, or some other from the respective range. You will have no competition there (for who didn't want to click the link, range is 198M), because most of other users are working either below 110M, or at 332M to hunt the EFF prize, therefore it will be no other user competing for it, most probably, so you can get that exponent as simple as using the manual test page, fill what exponent you want, or just add it to the worktodo.txt using a text editor.

[Petnek]

OK, where factor is found, there is no more work.

So I chose this one. Until when is reasonable to do TF? OK, this I found. For exponent up to 227,300,000 its 2^76 but here is next to GPU72 already 2^78. Why?
TF I can reserve for GPU, but for CPU I'm able to assign nothing. Why is that? I'm getting this error msg:

Error code: 40
Error text: No assignment available meeting CPU, program code and work preference requirements, cpu_id: 2233562, cpu # = 0, user_id = 28284

You wrote I can just add it to worktodo.txt. How I can get hash like part of the line written into that file?


I'm new in primes, so I'm not sure what is for what. Is here some sequence of tests? For example first TF, than P-1, PRP, ECM and finally LL? OK, this I found too. TF > P-1 > LL > DC. For what is PRP and ECM than?

[LaurV]

That range is TFed very low. As I said, nobody is (yet) interested in that range. 81 bits is a fair number for that range, and a good GPU. With an average GPU you can even go to 82 bits (average GPUs are weaker at LL/PRP, therefore they should do more TF).

Do you have a GPU? If you do, and if you want to do TF by yourself , then go to this site and click on your GPU. There will be a table opening at the beginning of the page, showing you how high you should TF.

There are tutorials on this forum about how to install mfaktX and go TF-ing (you need mfaktc for Nvidia GPUs and mfakto for AMD or opencl GPUs).

If you do not have a GPU or don't want to TF by yourself, then you should ask somebody to do TF for you. Don't do TF on CPU, it will be a waste of time. If you are serious on LL/PRP-ing M198705079, then I can do TF for you. Then you can do P-1 with CPU or again, ask somebody to do it for you. We have some threads here where people help each other. But you have to think seriously if you want to LL/PRP that exponent, it will take a long time (depending on your resources, from days to weeks, or even months).

For the sequence of tests, read the "math" page on Gimps for a start. The table shown there is from before the GPU era, when the TF was done on CPU. Entering the GPUs into the fray, they can factor hundreds of times faster, therefore, depending on your GPU, 5 to 10 bits can be added to that table. Each "bit" doubles the time spent factoring, so if your GPU can factor 10 times faster than your CPU, 3 bits are added (2^3=8), and if it can factor 500 times faster than your CPU, then 9 bits should be added (2^9=512).

You goth the order right. PRP is "new fashion" of LL and it may substitute it in time.

ECM is to find large factors of small mersenne numbers, not useful for the project (i.e. finding primes) but some people like to find those small gems called factors. Diamond rings are neither useful for anything (beside marking the level on the whiskey bottle).

[Petnek]

Yes, I have a GPU and I'm already TFing. In these times not so fast GTX960

The sequence of tests is clear to me, I hope

When or how is possible to assign P-1 test to that exponent? When I can do that, is it related to not enough TF tests or I'm somehow not reaching some requirments for categories? I'm getting that error msg mentioned in previous post.

[kriesel]

Welcome to the forum.
You might find some of the background or reference info at https://mersenneforum.org/showthread.php?t=24607 useful.
Historically LL was used for primality testing. Because of the excellent error detection of the Gerbicz error check, PRP is recommended instead of LL. Assuming no factors are found, the order is typicallyTF to the level merited by effort versus odds of finding a factor, P-1 factoring similarly, PRP and eventually PRP DC.

[LaurV]

(I edited my post above, crosspost, to add few phrases)

There should be no issue in getting a manual assignment in that range, for either P-1 or LL/PRP. You just go to gimps manual assignments, fill P-1 and the range close to your exponent. I just got this right now (which I am not going to do, most probably I will unreserve, but just to test the reservation form):

PFactor=33360C59858ED75AF8680544E5387FA8,1,2,198000067,-1,76,2

You can also add such lines manually to your worktodo, and let P95 communicate with the server, to record them (use N/A for the assignment key when you add manual work).
The P-1 work should be done AFTER the TF, or (depending on your system) before the last bit of TF, otherwise the work is in vain in case TF would have found a factor faster. So, if you decide to TF to 82 bits, the sequence should be TF to 81, then P-1 for about 3% or 5% chances to a factor, then TF to 82, then PRP (or LL, but right now, LL is frown here around, people want to switch to PRP for newer assignments, due to better "tolerance" for errors).

[Petnek]

Originally Posted by kriesel

Welcome to the forum.
You might find some of the background or reference info at https://mersenneforum.org/showthread.php?t=24607 useful.
Historically LL was used for primality testing...

Thank you for the welcome. I'm just going through all of these threads, interesting informations indeed.
OK, PRP is better than LL.

Originally Posted by LaurV

There should be no issue in getting a manual assignment in that range

I'm getting that error msg when trying to assign that exact exponent, no range around. That is not good way? How to assign P-1 just to chosen exponent?

Thank you both for further explanation of sequence of tests. I'm just getting through all those reccomended threads, so many info...

[Prime95]

Getting a PRP assignment implies that you have a P-1 assignment too. Both prime95 and the latest gpuOwl will do the P-1 automatically before running the PRP test.

[Petnek]

Thanks for all your answers.

So far I didn't find answer for this one... For which tests is wise to use GPU? Are GPUs faster on something else than TFing?

In my case I have GTX960, Ryzen 5 2600, i5-2410M and partially i7-6820HQ

[Uncwilly]

In general:
GPU for TF, it is almost like they were designed for this.
P-1 is good on a CPU
Primality checks (PRP or LL) do on a CPU.


This will best help the project along.
As always do what gives you pleasure.

There are times that using a GPU to do a first time check is warranted (as a platform independent check of a reported prime).