OFFICIAL "SERVER PROBLEMS" THREAD

[James Heinrich]

Originally Posted by kriesel

There is a clash between the maximum bit level that manual tf will allow (85), and the GPU72 listed final bit level at mersenne.ca (86) for high exponents in the mersenne.org range.

I wouldn't worry about it. TF levels are still being re-evaluated (with regards to PRP-proof rebalancing), and by the time the 990M range becomes relevant I'm sure things will have changed half a dozen ways. Target TF may drop by one bitlevel in the near(ish) future. I'm perfectly content to leave the mersenne.org pages capped at 85. If somebody really wants to TF to such depths I'm sure they'll figure out a way.

[kriesel]

Originally Posted by James Heinrich

I wouldn't worry about it. TF levels are still being re-evaluated (with regards to PRP-proof rebalancing), and by the time the 990M range becomes relevant I'm sure things will have changed half a dozen ways. Target TF may drop by one bitlevel in the near(ish) future. I'm perfectly content to leave the mersenne.org pages capped at 85. If somebody really wants to TF to such depths I'm sure they'll figure out a way.

Not worried, just thought it was interesting. It seems to only affect from ~956M up. It's easy enough to manually edit the last digit of a manual assignment. And performing these will be rare for quite some time.

[frmky]

On my assignments page at https://www.mersenne.org/workload/ I have two Manual testing PRP assignments from PRP-DC via gpuOwl that say they are original and current cat 0 and now expire in 3 days. However, when I go to https://www.mersenne.org/manual_extension/ to extend them, that table says they have an age of 4 days and expire in 176 days. Which is correct? Should I extend them? They will take about 4 more days to complete.

[Prime95]

Originally Posted by frmky

On my assignments page at https://www.mersenne.org/workload/ I have two Manual testing PRP assignments from PRP-DC via gpuOwl that say they are original and current cat 0 and now expire in 3 days. However, when I go to https://www.mersenne.org/manual_extension/ to extend them, that table says they have an age of 4 days and expire in 176 days. Which is correct? Should I extend them? They will take about 4 more days to complete.

I would trust the workload page. I'd extend them and hope it works.

[frmky]

Done. The extension page now says 236 days but the workload page still says 3 days. It's for 83877517 and 83878189.

[Prime95]

Originally Posted by frmky

Done. The extension page now says 236 days but the workload page still says 3 days. It's for 83877517 and 83878189.

I guess the workload page is thinking those are first-time tests whereas they are DCs.

[kriesel]

Originally Posted by Prime95

I guess the workload page is thinking those are first-time tests whereas they are DCs.

A long outstanding issue. Separate thread started in January: https://www.mersenneforum.org/showthread.php?t=25066
November: https://www.mersenneforum.org/showthread.php?t=24942

My workload page shows this assignment expiring in -109 days and is indicated as cat 0 there:
https://www.mersenne.org/report_expo...3023643&full=1
And the extensions page says it has 60 days before expiration.
According to assignment rules, it should be a CAT 4 PRP DC, and so presumably display and be treated as one.

(Please don't poach DC, especially of active reliable users; https://www.mersenne.org/report_expo...exp_hi=&full=1 was nearly 35% complete, and next in line to be resumed after completion in less than one more week, of two unintended 54M LLDC forced on it from V30.x updates. There's plenty of available LLDC or PRP DC to do for everyone, no need or justification to poach PRP DC assignments tens of millions away from the DC wavefront, costing/wasting system-weeks or months per poached exponent of a slow i3. On the laptop posting this message that's used daily. That one case 82972553 represents 2 months wasted on the laptop. PRP DC work type was selected both so that the slow system would remain far ahead of any wavefront, and aid in the accumulation of error rate data for PRP/GEC.)
https://www.mersenne.org/report_prp/...exfactor=1&B1= will return lots of potential DC. Those I spot checked would not involve poaching.

[kriesel]

Excluding doublecheck done, gives a list of active assignments, led off by PRP DC with pending/overdue doublecheck assignments (NOT expired, many are checking in regularly.) https://www.mersenne.org/assignments...&exp1=1&extf=1

[S485122]

On the Exponent Status Distribution page the numbers don't tally for the 3M, 4M, 6M, 7M and 10M ranges. (The number of exponents is not equal to the total of the number of primes, composites and unproven.) There is even an exponent that would be unproven : it has a 1 in the "Unproven / ERR" column, although all exponents that haven't been factored have been at least double checked.

Jacob

[lycorn]

Correct. I had also noticed that. No action yet, tho.

[Prime95]

Originally Posted by S485122

On the Exponent Status Distribution page the numbers don't tally for the 3M, 4M, 6M, 7M and 10M ranges. (The number of exponents is not equal to the total of the number of primes, composites and unproven.) There is even an exponent that would be unproven : it has a 1 in the "Unproven / ERR" column, although all exponents that haven't been factored have been at least double checked.

Jacob

Blame it on PRP proofs. Before, exponents were in one of a several tables. Now, while waiting for a proof to be uploaded the exponent is in none of those tables. It will get patched up someday.