Prime95 30.8 (big P-1 changes, see post #551)

[pepi37]

So only thing I changed from before is that 64*920^n+1 il a lot smaller then 3.8 M other candidates. Maybe B2 is depend on candidate size.
I will rerun test with TF value

https://www.mersenneforum.org/showpost.php?p=620989&postcount=812

This is how I work with Prime95 before: and one worker is always use full memory, second calculate first part....

[storm5510]

Originally Posted by pepi37

So only thing I changed from before is that 64*920^n+1 il a lot smaller then 3.8 M other candidates. Maybe B2 is depend on candidate size.
I will rerun test with TF value

https://www.mersenneforum.org/showpost.php?p=620989&postcount=812

This is how I work with Prime95 before: and one worker is always use full memory, second calculate first part....

How do you present this to Prime95 in the form you used above?

[pepi37]

Pminus1=64,920,308410,1,100000,20000000

[pepi37]

My question had bad formulation, so I will try to ask again. All files are in one zip, attached.
There is local.txt, results.txt, worktodo.txt and window.txt ( contain one worker window)
So question is ( for example) in two candidates in the row one use more memory, and second use far less memory, even fact they have nearly 29 GB RAM free.
So what algo Prime95 use, and how I can say that Prime95 use all memory for every test.

[May 19 16:17:55] Available memory is 28594MB.
[May 19 16:17:55] Switching to all-complex FMA3 FFT length 640K, Pass1=512, Pass2=1280, clm=2, 4 threads
[May 19 16:17:55] Estimated stage 2 vs. stage 1 runtime ratio: 0.765
[May 19 16:17:57] Using 22057MB of memory.  D: 1050, 120x4184 polynomial multiplication.
[May 19 16:18:00] Stage 2 init complete. 2948 transforms. Time: 4.460 sec.
[May 19 16:18:51] 64*920^462706+1 stage 2 complete. 95619 transforms. Total time: 50.533 sec.
[May 19 16:18:51] Round off: 0.34375
[May 19 16:18:51] Stage 2 GCD complete. Time: 0.643 sec.
[May 19 16:18:51] 64*920^462706+1 completed P-1, B1=100000, B2=13665750, Wi4: 0CCC1D93
[May 19 16:18:51] 
[May 19 16:18:51] P-1 on 64*920^463018+1 with B1=100000, B2=10000000
[May 19 16:18:51] Chance of finding a factor is an estimated 4.62%
[May 19 16:18:51] Using all-complex FMA3 FFT length 512K, Pass1=512, Pass2=1K, clm=1, 4 threads
[May 19 16:19:59] 64*920^463018+1 stage 1 complete. 288858 transforms. Total time: 67.858 sec.
[May 19 16:19:59] Round off: 0.2109375
[May 19 16:20:00] Inversion of stage 1 result complete. 5 transforms, 1 modular inverse. Time: 1.217 sec.
[May 19 16:20:01] Available memory is 28594MB.
[May 19 16:20:01] Switching to all-complex FMA3 FFT length 640K, Pass1=512, Pass2=1280, clm=2, 4 threads
[May 19 16:20:01] Estimated stage 2 vs. stage 1 runtime ratio: 0.735
[May 19 16:20:02] Using 16582MB of memory.  D: 1050, 120x3120 polynomial multiplication.
[May 19 16:20:05] Stage 2 init complete. 2948 transforms. Time: 4.463 sec.
[May 19 16:20:52] 64*920^463018+1 stage 2 complete. 93121 transforms. Total time: 47.326 sec.
[May 19 16:20:52] Round off: 0.3125
[May 19 16:20:53] Stage 2 GCD complete. Time: 0.631 sec.
[May 19 16:20:53] 64*920^463018+1 completed P-1, B1=100000, B2=13304550, Wi4: 0CC39868

[storm5510]

Originally Posted by pepi37

My question had bad formulation, so I will try to ask again. All files are in one zip, attached.
There is local.txt, results.txt, worktodo.txt and window.txt ( contain one worker window)
So question is ( for example) in two candidates in the row one use more memory, and second use far less memory, even fact they have nearly 29 GB RAM free.
So what algo Prime95 use, and how I can say that Prime95 use all memory for every test.

In local.txt, you have it set to:

Memory=28600 during 7:30-23:30 else 29000

Prime95 may not use all of what you have allowed. Only what it needs. I do not see a problem with what you have, but someone else might.

[pepi37]

Originally Posted by storm5510

In local.txt, you have it set to:

Memory=28600 during 7:30-23:30 else 29000

Prime95 may not use all of what you have allowed. Only what it needs. I do not see a problem with what you have, but someone else might.

If you read my zip file you will find that is set already
Memory=28600 during 7:30-23:30 else 29000

Problem is next: when Prime95 use more memory then will go further in B2 and change for find factor will be better.
Maybe I need to lift up B1 bound?

[Prime95]

Originally Posted by pepi37

If you read my zip file you will find that is set already
Memory=28600 during 7:30-23:30 else 29000

This seems silly. What this ensures is that P-1 will needlessly stop and restart for new memory settings. Just set it to 28600 all day (or 29000).

Problem is next: when Prime95 use more memory then will go further in B2 and change for find factor will be better. Maybe I need to lift up B1 bound?

Prime95 needs to know how much TF has been done for it to pick the near optimal B2 bounds. Otherwise, prime95 will use the B2 value from worktodo.txt

[pepi37]

Originally Posted by Prime95

This seems silly. What this ensures is that P-1 will needlessly stop and restart for new memory settings. Just set it to 28600 all day (or 29000).



Prime95 needs to know how much TF has been done for it to pick the near optimal B2 bounds. Otherwise, prime95 will use the B2 value from worktodo.txt

Is any way to force all memory not optimal b2 bound? As you suggest I set Tf bound in woroktodo.txt. Also will upper b1 produce better chance of finding factor?

[storm5510]

Originally Posted by pepi37

Is any way to force all memory not optimal b2 bound? As you suggest I set Tf bound in woroktodo.txt. Also will upper b1 produce better chance of finding factor?

Match your memory settings!

You seem to be fighting with what you are doing. B1 uses very little memory. B2 will adjust itself based on your memory setting if you allow it. You are trying to force it using the Pminus1 form. I do not believe this will do any good. It may even slow it down. I feel it would be better to use the PFactor form. This will allow Prime95 to use what it needs, as I described above.

[pepi37]

Originally Posted by storm5510

Match your memory settings!

You seem to be fighting with what you are doing. B1 uses very little memory. B2 will adjust itself based on your memory setting if you allow it. You are trying to force it using the Pminus1 form. I do not believe this will do any good. It may even slow it down. I feel it would be better to use the PFactor form. This will allow Prime95 to use what it needs, as I described above.

If I use your solution with Pfactor then I got this

64*920^300082+1 does not need P-1 factoring.

So this is not solution (for my case)

[pepi37]

Originally Posted by storm5510

Match your memory settings!

You seem to be fighting with what you are doing. B1 uses very little memory. B2 will adjust itself based on your memory setting if you allow it. You are trying to force it using the Pminus1 form. I do not believe this will do any good. It may even slow it down. I feel it would be better to use the PFactor form. This will allow Prime95 to use what it needs, as I described above.

Or better to say / kindly ask please write one line what I should put in worktodo.txt to get benefit from using Pfactor. You have all info you need to know. Thanks in advance