processed dc and tc posts

[kriesel]

Cat 1
Doublecheck=55744531,74,1 done
Doublecheck=55754381,74,1 done

The following are reserved, running, and expected done by 3/15/21
Doublecheck=55820617,74,1
Doublecheck=55847947,74,1
Doublecheck=55857943,74,1
Doublecheck=55882187,74,1
Doublecheck=55894987,74,1
Doublecheck=55953253,74,1

Cat3
Doublecheck=59876611,74,1

[endless mike]

Originally Posted by Uncwilly

List updated. There was one exponent that had a suspect result and 2 non-matching non-suspect results. I am running that one off books as PrimeNet won't let me have it.

I saw that result come in and match. Thanks!

[kriesel]

cat 1
55971199 done
55972223 done
55986647 in progress
55993999 in progress
56026717 in progress
56083631 queued
56099531 in progress
56100299 queued

cat 2
56453669 in progress
57184763 queued

projected all above done NLT Friday 3/19/2021
(Onward to Mp#48!)

[Uncwilly]

Originally Posted by kriesel

projected all above done NLT Friday 3/19/2021
(Onward to Mp#48!)

Very good. The list has been updated

[kriesel]

Cat 1
Doublecheck=56142673,74,1
Doublecheck=56180083,74,1
Doublecheck=56206543,74,1
Doublecheck=56235139,74,1

Cat 2
Doublecheck=57034589,74,1

[kriesel]

Have you considered also screening 100Mdigit twice-tested exponents for residue mismatches?
I'm tackling 332194529, which is the first Mersenne exponent that was LL tested to completion without detected errors per the milestones page, (with two results, and unmatched residues). There does not seem to be any way to reserve the test manually, after trying several assignment types there. Nor did inserting DoubleCheck=332194529,-1,81,1 into a prime95 worktodo.txt and following the usual manual communication instructions to get PrimeNet to issue an AID. (Nor the equivalent for PRP.) It's probably a much better candidate for PRP with proof, but that assignment won't issue manually either. Hence the value in having it understood what is under way, by means other than the PrimeNet server for the moment.

Some other mismatches in 100M LL DC unresolved currently (1 suspect result in each case):
332197123
332250197 needs better P-1, which I'm running now
340705633

A rough gauge of error rate in 100M LL & DC is provided by the 17 exponents that are listed as verified at https://www.mersenne.org/report_ll/?...exfactor=1&B1=
in which there were 5 known-bad results after LL TC:
332197309
332227579
332229923
332980399
402637217
(5 bad of 39 tests on 17 exponents)
Plus 8 tests in the 4 I found mismatched without TC, showing at least 4 bad results there.

So, approx ~19% observed error rate, in 47 tests on 21 of the 100Mdigit exponents, at least 9 unmatched & presumably bad results, 9/47.
(That extrapolates to about 156 undiscovered bad 100Mdigit unverified LL first tests in the 814 listed by https://www.mersenne.org/report_ll/?...exfactor=1&B1=)

These are not suited for lightweight hardware. I see runtime estimates of 2 weeks on recent gpuowl and a Radeon VII, or 3.5 months in prime95 on a 12-core Xeon e5-2697v2.

PRP with mismatches:
332826083 type1/type4 clash
332897017 type1/type4 clash
These are candidates for type 1 PRP with proof generation

[Uncwilly]

Originally Posted by kriesel

Have you considered also screening 100Mdigit twice-tested exponents for residue mismatches?

Maybe in an organized way when the DC wave front gets up near 95M. Until that time there will be plenty of strategic work below that range. Nothing stopping anyone like you or Slaur from freelancing up there. The assignment rules should be set to not handout new LL work in that range. The PrimeNet server should give out a specific warning that exponents in that range need to be done via PRP with GEC and by software that can generate 'proof' files.

Nor did inserting DoubleCheck=332194529,-1,81,1 into a prime95 worktodo.txt and following the usual manual communication instructions to get PrimeNet to issue an AID.

Did you try it without the highlighted term?

There were a lot of errors around 33.2M as well.

[kriesel]

Originally Posted by Uncwilly

The assignment rules should be set to not handout new LL work in that range. The PrimeNet server should give out a specific warning that exponents in that range need to be done via PRP with GEC and by software that can generate 'proof' files.Did you try it without the highlighted term?

There were a lot of errors around 33.2M as well.

Or from 120M up, PRP/GEC required, except perhaps for DC.332194529 DC won't reassign, because it's still assigned to "HHfromG" since March 2014, with zero progress.

[ChZ]

Requesting LL triple-check:

DoubleCheck=64227263,76,1

[kriesel]

Following are barring any system disruptions:
56100299 updated projection 3/21/21

Cat 1 in progress est Mar 23
Doublecheck=56068889,74,1
Doublecheck=56216921,74,1
Doublecheck=56247413,74,1
Doublecheck=56257001,74,1
Doublecheck=56268127,74,1


Doublecheck=64227263,76,1 projected 9 hours

[kriesel]

Following are reserved, and barring any system disruptions, ETA Mar 26 or 27:

Cat 2
Doublecheck=58167337,74,1
Doublecheck=58862051,74,1
Doublecheck=59107187,74,1

Exponents with 1 Suspect result + 1 Unverified result:
Doublecheck=58587167,74,1
Doublecheck=58682489,74,1
Doublecheck=59908019,74,1