mersenneforum.org Why can't I get a manual PRP test of M103747177 ?
 Register FAQ Search Today's Posts Mark Forums Read

 2021-05-05, 10:19 #1 drkirkby   "David Kirkby" Jan 2021 Althorne, Essex, UK 5×37 Posts Why can't I get a manual PRP test of M103747177 ? I wanted to check the memory usage of a P-1 test, as it was implied on another thread that giving P-1 tests extra RAM would allow it to find bigger factors. So I set the RAM limit to 170 GB in mprime, reserved a P-1 test, and got assigned M103747177. (mprime is not using much RAM, but that's not my main concern) Nobody, other than myself (David Kirkby), is running any tests on that exponent, https://www.mersenne.org/report_expo...exp_hi=&full=1 and since it is fairly small on the scale of world record Mersenne primes, I would like to do a PRP test on it. So I manually requested a PRP test, putting that as the upper and lower bounds on the PRP test. But the Primenet server reports nothing is available, despite nobody else is doing a PRP test. After putting that as a lower limit, and not putting an upper limit, I got allocated 110474729 as a PRP test. Obviously its going to take longer to test 110474729 than it would 103747177, so I would rather do the smaller exponent. . Is there a line I could put in worktodo.txt, after the P-1 test, so I get a PRP test of 103747177? If so, what should I put? So far it has been trial factored to 2^76. Of course the P-1 test might find a factor, in which case the PRP test would be pointless. Maybe that's why I can't get a PRP test, but can I force one if the P-1 test does not find a factor? Last fiddled with by drkirkby on 2021-05-05 at 10:19
 2021-05-05, 10:39 #2 kruoli     "Oliver" Sep 2017 Porta Westfalica, DE 7568 Posts According to this page, the exponent is in category 1 and thus not available for manual assignments. If you want to test the exponent, you'd have to own a machine that qualifies for that category, but I guess you dual Xeon machine will likely be that. Additionally, a exponent cannot be assigned for PRP while there is a P-1 running on it. You have to exit mprime and add PRP=1,2,103747177,-1,76,0,3,1 to worktodo.txt. If you do this after the P-1 completed, mprime will try and reserve that exponent, but be quick, it is likely that someone else might get it after the P-1 result is turned in. If you do not get the assignment, please don't run the PRP test. Last fiddled with by kruoli on 2021-05-05 at 10:40 Reason: Corrected worktodo.txt entry.
2021-05-05, 12:09   #3
drkirkby

"David Kirkby"
Jan 2021
Althorne, Essex, UK

18510 Posts

Quote:
 Originally Posted by kruoli According to this page, the exponent is in category 1 and thus not available for manual assignments. If you want to test the exponent, you'd have to own a machine that qualifies for that category, but I guess you dual Xeon machine will likely be that. Additionally, a exponent cannot be assigned for PRP while there is a P-1 running on it. You have to exit mprime and add PRP=1,2,103747177,-1,76,0,3,1 to worktodo.txt. If you do this after the P-1 completed, mprime will try and reserve that exponent, but be quick, it is likely that someone else might get it after the P-1 result is turned in. If you do not get the assignment, please don't run the PRP test.
Thank you. One problem I have with this 52-core Dell 7920 is that it had a lot of old exponents which were started on slow machines I had access for a short period of time. The GIMPS server keeps thinking exponents are going to take ages, when in fact one PRP test is completed each day. In an attempt to get around this, I have unreserved lots of old exponents, except the two which had more than 2% compoleted. I have received newer exponents today. Hopefully the server will realise I can complete exponents quickly.

My real interest is PRP tests, but I'm wondering if completing a few P-1 tests quickly might make the server realise the Dell is not as slow as it thinks, so give me some smaller PRP exponents. I don't know how long this P-1 test will take, but I'm guessing around another 2 hours.
Code:
[Worker #1 May 5 10:17] Optimal P-1 factoring of M103747177 using up to 286720MB of memory.
[Worker #1 May 5 10:20] M103747177 stage 1 is 9.05% complete. Time: 203.402 sec.
[Worker #1 May 5 10:24] M103747177 stage 1 is 18.12% complete. Time: 202.782 sec.
[Worker #1 May 5 10:27] M103747177 stage 1 is 27.18% complete. Time: 202.729 sec.
[Worker #1 May 5 10:30] M103747177 stage 1 is 36.24% complete. Time: 202.749 sec.
[Worker #1 May 5 10:34] M103747177 stage 1 is 45.31% complete. Time: 202.779 sec.
[Worker #1 May 5 10:37] M103747177 stage 1 is 54.37% complete. Time: 202.526 sec.
[Worker #1 May 5 10:40] M103747177 stage 1 is 63.43% complete. Time: 202.620 sec.
[Worker #1 May 5 10:44] M103747177 stage 1 is 72.49% complete. Time: 202.571 sec.
[Worker #1 May 5 10:47] M103747177 stage 1 is 81.56% complete. Time: 202.447 sec.
[Worker #1 May 5 10:51] M103747177 stage 1 is 90.62% complete. Time: 202.557 sec.
[Worker #1 May 5 10:54] M103747177 stage 1 is 99.68% complete. Time: 202.224 sec.
[Worker #1 May 5 10:54] M103747177 stage 1 complete. 2206916 transforms. Time: 2236.852 sec.
[Worker #1 May 5 11:12] M103747177 stage 2 is 5.80% complete. Time: 1059.691 sec.
[Worker #1 May 5 11:30] M103747177 stage 2 is 14.37% complete. Time: 1073.976 sec.
[Worker #1 May 5 11:48] M103747177 stage 2 is 22.95% complete. Time: 1073.885 sec.
[Worker #1 May 5 12:15] Optimal P-1 factoring of M103747177 using up to 286720MB of memory.
[Comm thread May 5 12:15] Sending expected completion date for M103747177: May  5 2021
[Worker #1 May 5 12:16] Optimal P-1 factoring of M103747177 using up to 286720MB of memory.
[Worker #1 May 5 12:17] Optimal P-1 factoring of M103747177 using up to 286720MB of memory.
[Worker #1 May 5 12:17] Optimal P-1 factoring of M103747177 using up to 286720MB of memory.
[Worker #1 May 5 12:18] Optimal P-1 factoring of M103747177 using up to 286720MB of memory.
[Worker #1 May 5 12:21] Optimal P-1 factoring of M103747177 using up to 286720MB of memory.
[Worker #1 May 5 12:27] M103747177 stage 2 is 28.85% complete.
[Worker #1 May 5 12:45] M103747177 stage 2 is 37.42% complete. Time: 1072.443 sec.
[Worker #1 May 5 13:02] M103747177 stage 2 is 45.99% complete. Time: 1072.253 sec.

Last fiddled with by drkirkby on 2021-05-05 at 12:13

2021-05-05, 14:55   #4
drkirkby

"David Kirkby"
Jan 2021
Althorne, Essex, UK

101110012 Posts

Quote:
 Originally Posted by kruoli According to this page, the exponent is in category 1 and thus not available for manual assignments. If you want to test the exponent, you'd have to own a machine that qualifies for that category, but I guess you dual Xeon machine will likely be that. Additionally, a exponent cannot be assigned for PRP while there is a P-1 running on it. You have to exit mprime and add PRP=1,2,103747177,-1,76,0,3,1 to worktodo.txt. If you do this after the P-1 completed, mprime will try and reserve that exponent, but be quick, it is likely that someone else might get it after the P-1 result is turned in. If you do not get the assignment, please don't run the PRP test.
I got it assigned to me
https://www.mersenne.org/report_expo...exp_hi=&full=1
as a category 1 assignment, despite the PRP test had probably finished 20 minutes earlier.

The P-1 tested used 47 GB RAM and the website shows "B1=765000, B2=22708000, E=12"

Hopefully, if I can sort out why the server thinks my Dell is so slow, I will automatically get some category 0 and 1 assignments, as I can do them in two days on two CPUs, so complete one per day.

Dave

Last fiddled with by drkirkby on 2021-05-05 at 14:56

2021-05-05, 15:01   #5
kruoli

"Oliver"
Sep 2017
Porta Westfalica, DE

2·13·19 Posts

Quote:
 Originally Posted by drkirkby The P-1 tested used 47 GB RAM [...]
According to this calculator, that amount of RAM you observed is actually the maximum of what mprime will use on an exponent of this size. So any memory above that will not be used. For higher exponents (e.g. 333000001), more memory might be used (in this example up to 143 GB).

2021-05-05, 15:38   #6
S485122

Sep 2006
Brussels, Belgium

2×5×167 Posts

Quote:
 Originally Posted by drkirkby ... Hopefully, if I can sort out why the server thinks my Dell is so slow, I will automatically get some category 0 and 1 assignments, as I can do them in two days on two CPUs, so complete one per day. Dave
The server doesn't think your machine is slow.

To be able to get something else than CAT 4 assignments one must meet some conditions, read the relevant page : one must register the computer to get the smallest assignments, the computer must have returned enough good results in a certain period of time, it must not request too many days of work in advance...

The only speed requirement is that the assignment will be completed in 30 days.

Jacob

2021-05-05, 17:45   #7
axn

Jun 2003

32·19·29 Posts

Quote:
 Originally Posted by drkirkby The P-1 tested used 47 GB RAM and the website shows "B1=765000, B2=22708000, E=12"
Quote:
 Originally Posted by kruoli According to this calculator, that amount of RAM you observed is actually the maximum of what mprime will use on an exponent of this size. So any memory above that will not be used.
You should upgrade your client to version >= 30.4 (latest is 30.6), as those have improved (i.e. much faster) stage 2 and which will use all the memory you can throw at it.

2021-05-05, 19:21   #8
drkirkby

"David Kirkby"
Jan 2021
Althorne, Essex, UK

5·37 Posts

Quote:
 Originally Posted by axn You should upgrade your client to version >= 30.4 (latest is 30.6), as those have improved (i.e. much faster) stage 2 and which will use all the memory you can throw at it.
Code:
[dkirkby@jackdaw tmp]\$ ./mprime -v
Mersenne Prime Test Program: Linux64,Prime95,v30.3,build 6
It crossed my mind later, that it is probably unwise to allocate more than half the memory to P-1 factoring on a dual processor machine, because each CPU has its own memory .Whilst CPUs can access the memory of another CPU, that is slower.
I think discussions about the RAM usage are probably best on a software forum, as its not related to the server.

Dave

Last fiddled with by drkirkby on 2021-05-05 at 19:22

2021-05-05, 19:39   #9
drkirkby

"David Kirkby"
Jan 2021
Althorne, Essex, UK

5×37 Posts

Quote:
 Originally Posted by S485122 The server doesn't think your machine is slow. To be able to get something else than CAT 4 assignments one must meet some conditions, read the relevant page : one must register the computer to get the smallest assignments, the computer must have returned enough good results in a certain period of time, it must not request too many days of work in advance... The only speed requirement is that the assignment will be completed in 30 days. Jacob
I thought had configured the website to give me the smallest exponents, but I had not. I could have swore I did that before, but maybe I am going mad.

I adjusted the days of work to queue to be 3 now (DaysOfWork=3). I'm puzzled that the server gives me several exponents it thinks are going to take up to 42 days to complete. Currently it thinks it will get exponents in 3, 10, 13, 20, 31 and 42 days. In practice, the lot should be done in about 6 days, as I can do about one per day. But even 6 days is twice the setting of DaysOfWork, which is set to 3.

Dave

Last fiddled with by drkirkby on 2021-05-05 at 19:39

 2021-05-05, 19:47 #10 Uncwilly 6809 > 6502     """"""""""""""""""" Aug 2003 101×103 Posts 2×34×59 Posts What is your "Estimated hours per day this program will run" under Options->CPU?
2021-05-05, 19:55   #11
drkirkby

"David Kirkby"
Jan 2021
Althorne, Essex, UK

5×37 Posts

Quote:
 Originally Posted by Uncwilly What is your "Estimated hours per day this program will run" under Options->CPU?
24 and it is pretty much running 24/7.

I think the issue might be I had some exponents on it that were allocated a month or more ago on slower machines, which I moved to this machine. I have now unassigned all them - most had been started, but less than 1% done, so no real wastage.

But the server assigned me several exponents today, despite I had set the days of work to queue to be 3. I have 7 exponents on my Dell, which will take me about 7 days to complete. So why did the server assign them, if it thinks they will take up to 42 days to complete?

Last fiddled with by drkirkby on 2021-05-05 at 19:58

 Similar Threads Thread Thread Starter Forum Replies Last Post sweety439 sweety439 7 2020-02-11 19:49 Trilo Miscellaneous Math 25 2018-03-11 23:20 lidocorc Software 3 2008-12-03 15:12 GrK GMP-ECM 2 2007-12-25 01:55 T.Rex Math 0 2004-10-26 21:37

All times are UTC. The time now is 22:34.

Sat May 8 22:34:10 UTC 2021 up 30 days, 17:15, 0 users, load averages: 3.20, 3.00, 3.13