mersenneforum.org Should I re-do PM1?
 Register FAQ Search Today's Posts Mark Forums Read

 2021-06-15, 01:52 #1 JuanTutors     "Juan Tutors" Mar 2004 13·43 Posts Should I re-do PM1? I am working on a 100M digit exponent and I noticed that the P-1 test has been done with bounds B1=100000, B2=1000000. Even on a low-ram PC, I blow these bounds out of the water. In terms of saving a PRP test, should I re-do the PM1 test?
 2021-06-15, 03:00 #2 kriesel     "TF79LL86GIMPS96gpu17" Mar 2017 US midwest 19×311 Posts Go for the GPU72 bounds or at least the PrimeNet bounds https://www.mersenne.ca/exponent/332194529 for example or use the calculator at https://www.mersenne.ca/prob.php?exp...&factorbits=81 No half-measures first. Go for adequate bounds first time around, so it's one and done. Anything less is a waste. https://www.mersenneforum.org/showpo...9&postcount=20 Last fiddled with by kriesel on 2021-06-15 at 03:01
 2021-06-15, 03:14 #3 Uncwilly 6809 > 6502     """"""""""""""""""" Aug 2003 101×103 Posts 2×33×11×17 Posts I concur.
 2021-06-15, 05:33 #4 JuanTutors     "Juan Tutors" Mar 2004 13·43 Posts How many tests are saved with those bounds? I set the number of tests saved to 1, 2, and 1.1 just to experiment and I can't reproduce those bounds.
2021-06-15, 10:11   #5
axn

Jun 2003

10100001110112 Posts

Quote:
 Originally Posted by JuanTutors How many tests are saved with those bounds? I set the number of tests saved to 1, 2, and 1.1 just to experiment and I can't reproduce those bounds.
Don't worry about reproducing the outcome of the calculator. Just set tests saved to 1, give as much memory as you can spare, and let P95 / gpuowl calculate the optimal bounds.

2021-06-15, 15:16   #6
JuanTutors

"Juan Tutors"
Mar 2004

10578 Posts

Quote:
 Originally Posted by axn Don't worry about reproducing the outcome of the calculator. Just set tests saved to 1, give as much memory as you can spare, and let P95 / gpuowl calculate the optimal bounds.
OK that's what I'm doing then, although I set tests saved to 1.01 to potentially save the DC. (I tried setting the tests saved to 1.005 since the website says the DC takes less than 0.5% time, but tests saved only takes decimals down to the hundredths place.) The bounds are B1=1309000 and B2=60901000.

Last fiddled with by JuanTutors on 2021-06-15 at 15:22

 2021-06-15, 17:39 #7 axn     Jun 2003 5,179 Posts Sounds reasonable. With the older "2 tests saved" setting it would've spent more than 2x that amount.
 2021-06-16, 19:42 #8 JuanTutors     "Juan Tutors" Mar 2004 10578 Posts About the trial factoring to GPU72 depth, I don't have a way to TF the exponent to that depth. I tried setting it up on my computers and failed miserably. Is there a thread to make TF requests for exponents that are currently being P-1 tested? Last fiddled with by JuanTutors on 2021-06-16 at 19:42
 2021-06-16, 20:10 #9 Uncwilly 6809 > 6502     """"""""""""""""""" Aug 2003 101×103 Posts 2×33×11×17 Posts Have you tried doing a google Colab? There is material about it in the GPU72 section. If you haven't used it you could get the TF done. Somewhere in one of the threads Chalsall posted a bit of code to all you to add in an assignment of your choice.
2021-06-18, 01:04   #10
JuanTutors

"Juan Tutors"
Mar 2004

13·43 Posts

Quote:
 Originally Posted by Uncwilly Have you tried doing a google Colab? There is material about it in the GPU72 section. If you haven't used it you could get the TF done. Somewhere in one of the threads Chalsall posted a bit of code to all you to add in an assignment of your choice.
That's new to me. I tried to get the instructions between students. Hopefully I can get that going. I will try again tomorrow.

2021-06-29, 17:00   #11
JuanTutors

"Juan Tutors"
Mar 2004

10578 Posts

Quote:
 Originally Posted by Uncwilly Have you tried doing a google Colab? There is material about it in the GPU72 section. If you haven't used it you could get the TF done. Somewhere in one of the threads Chalsall posted a bit of code to all you to add in an assignment of your choice.
I found a few threads about Google Colab. Is there a step-by-step instruction manual on using Google Colab for TF? Like paste this here, change this here, etc?

All times are UTC. The time now is 06:40.

Sun Nov 28 06:40:01 UTC 2021 up 128 days, 1:09, 0 users, load averages: 0.64, 0.83, 0.91