Strategic P-1 for DC (requires 24GB or more memory for prime95)

[Prime95]

Originally Posted by Rubiksmath

I've had a roundoff error on 30.9 ECM a few times (extra roundoff checking not selected so it was p95 itself getting concerned). It was on an exponent right around M214xxx where it the bits-per-word was 0.009 bits below the maximum (extremely slim margin) and at about 90% stage 1 it threw a round off error of around 0.45 which caused it to abort the run and restart with higher FFT length.

Originally Posted by lycorn

I´ve noticed the same. Extra roundoff checking not selected either. Exponents (two, so far) in the low 800k range. That was what prompted my last post about ECM concerns, as I am currently doing some extensive ECM testing in this range.

Please email/PM me the details whenever this happens. I'll need the ECM exponent and machine architecture (AVX-512, FMA, AVX, or SSE2). With that info I can change the default FFT crossovers for a future release.

[Prime95]

Originally Posted by Prime95

Please use version 30.8. There is a bug in version 30.9 - 30.12b1.

Preliminary news.

Further investigation indicates the problem is not where I thought it would be. Version 30.9 (all builds) are OK. The bug does not affect ECM. If exponent divided by stage-2-FFT-length is greater than about 16.1 then there is a good chance the bug has reared its ugly head.

If my preliminary analysis is correct, then about 450 P-1 results are affected.

[kriesel]

+48:
Pfactor=1,2,101940353,-1,77,1
to
Pfactor=1,2,101943913,-1,77,1

But, these were already assigned:
;Pfactor=N/A,1,2,101940533,-1,77,1
;Pfactor=N/A,1,2,101941813,-1,77,1
;Pfactor=N/A,1,2,101941823,-1,77,1
;Pfactor=N/A,1,2,101942017,-1,77,1
;Pfactor=N/A,1,2,101942261,-1,77,1
;Pfactor=N/A,1,2,101942381,-1,78,1
;Pfactor=N/A,1,2,101942413,-1,77,1
;Pfactor=N/A,1,2,101942447,-1,78,1
;Pfactor=N/A,1,2,101942653,-1,77,1
;Pfactor=N/A,1,2,101943913,-1,77,1

[kriesel]

+40:
Pfactor=1,2,101943953,-1,77,1
to
Pfactor=1,2,101957923,-1,77,1

But, already assigned:
;Pfactor=N/A,1,2,101943953,-1,77,1
;Pfactor=N/A,1,2,101943971,-1,77,1
;Pfactor=N/A,1,2,101944463,-1,77,1

[kriesel]

+40:
Pfactor=1,2,101957981,-1,77,1
to
Pfactor=1,2,101970391,-1,78,1
But these were already assigned:
;Pfactor=N/A,1,2,101960087,-1,77,1
;Pfactor=N/A,1,2,101960107,-1,78,1

[kriesel]

+40:
Pfactor=1,2,101970689,-1,77,1
to
Pfactor=1,2,101986697,-1,77,1

But the following 13 (32.5%!) were found at initial manual communication to be already assigned elsewhere, so were manually removed from worktodo files:
Pfactor=N/A,1,2,101982719,-1,77,1
Pfactor=N/A,1,2,101982781,-1,78,1
Pfactor=N/A,1,2,101982883,-1,77,1
Pfactor=N/A,1,2,101983073,-1,77,1
Pfactor=N/A,1,2,101983097,-1,77,1
Pfactor=N/A,1,2,101983129,-1,77,1
Pfactor=N/A,1,2,101983283,-1,77,1
Pfactor=N/A,1,2,101983303,-1,77,1
Pfactor=N/A,1,2,101983333,-1,77,1
Pfactor=N/A,1,2,101984059,-1,78,1
Pfactor=N/A,1,2,101984263,-1,77,1
Pfactor=N/A,1,2,101984549,-1,77,1
Pfactor=N/A,1,2,101986697,-1,77,1

[kriesel]

The last 30:
Pfactor=1,2,101986847,-1,77,1
to
Pfactor=1,2,101991053,-1,77,1

[kriesel]

+30:
Pfactor=1,2,105236503,-1,77,1
to
Pfactor=1,2,105492901,-1,78,1

[WillowW]

Hi all!

There seems to be a month-long gap in this thread, where posts have gone missing, from March 27th to April 22.

Perhaps it's normal maintenance, perhaps it's a glitch in the database? It's not as though the collected works of Shakespeare have been lost. ;) But I thought I'd mention the gap, because it seemed odd and might represent a more serious problem.

Warm hellos, Willow

[kruoli]

Originally Posted by WillowW

Perhaps it's normal maintenance, perhaps it's a glitch in the database?

Thankfully the former, but better ask that let it go undetected.

[S485122]

Originally Posted by WillowW

Hi all!

There seems to be a month-long gap in this thread, where posts have gone missing, from March 27th to April 22.

Perhaps it's normal maintenance, perhaps it's a glitch in the database? It's not as though the collected works of Shakespeare have been lost. ;) But I thought I'd mention the gap, because it seemed odd and might represent a more serious problem.

Warm hellos, Willow

I think that the reservation posts are periodically cleaned (deleted). Other posts might be kept.