OFFICIAL "SERVER PROBLEMS" THREAD

[Prime95]

Originally Posted by GP2

For PRP primality double-checking:

Obviously there is no list of known factors, and after the -1 there are the fields "75,0,3,4" (no factors below 2⁷⁵, uses base 3 as the work type, and uses residue type 4, to match the first time test that was done by preda using gpuOwL).

Early gpuOwl's did type 4, later versions do type 1.

[Madpoo]

Originally Posted by GP2

Well, not exactly.

For PRP cofactor double checking:

Between the -1 and the comma-separated list of known factors, there is 99,0,3,1 if it's a double-check of a Type 1 first-time result and 99,0,3,5 if it's a double check of a first-time type 5 result.

For PRP cofactor first-time checking:

It's still a valid line if those four fields are omitted, and I think it just defaults to Type 5. If I'm not mistaken, first-time PRP test assignments do omit those fields, but I don't have any current examples to look at.

For PRP primality double-checking:

Obviously there is no list of known factors, and after the -1 there are the fields "75,0,3,4" (no factors below 2⁷⁵, uses base 3 as the work type, and uses residue type 4, to match the first time test that was done by preda using gpuOwL).

For PRP primality first-time checking:

Obviously there is no list of known factors, and the fields just end with the -1

Oof. Just when I thought I was getting close to a basic test...

Now my alterations will need to do some additional work to see:
- is this a double-check? If so, set the residue type and base to the same as the first one
- also if it's a double-check, it would need to be using the same # of known factors, although if a new one had been found in the meanwhile, it may or may not be useful to do this double-check anyway (depending on the residue type)

It also depends on the client version since older clients won't support those extra fields (residue type, base, etc).

Well, at any rate I've just about got it worked up to do the basic stuff without the extra parts you mention.

[James Heinrich]

Originally Posted by Madpoo

It also depends on the client version since older clients won't support those extra fields (residue type, base, etc).

Should you even make an effort to consider such? If the client is old and doesn't support said fields, perhaps we don't want it running PRP tests anyways?

[Madpoo]

Originally Posted by Prime95

Early gpuOwl's did type 4, later versions do type 1.

What's the default type for the latest Prime95? I could omit it and let it do whatever (if this is the first PRP test) or specifically set it to whatever the default ... is it type 5?

I have it sort of mocked up but I really only tested it with a couple of PRP cofactor assignments on myself. I have every reason to believe there are still bugs in it. :) You may even get a 500 error or some crazy debug text...

https://www.mersenne.org/workload/default.mock.php

[potonono]

That looks good to me, though my Prime95 worktodo file doesn't include the last few delimited items.

mersenne.org/workload/default.mock.php shows:
PRP=xxxyyyzzz,1,2,mmmmmmmm,-1,75,0,3,1
PRP=aaabbbccc,1,2,mmmmmmmm,-1,75,0,3,5

Prime95 v29.4.5.0 worktodo.txt shows:
PRP=xxxyyyzzz,1,2,mmmmmmmm,-1
PRP=aaabbbccc,1,2,mmmmmmmm,-1

[Madpoo]

Originally Posted by potonono

That looks good to me, though my Prime95 worktodo file doesn't include the last few delimited items.

mersenne.org/workload/default.mock.php shows:
PRP=xxxyyyzzz,1,2,mmmmmmmm,-1,75,0,3,1
PRP=aaabbbccc,1,2,mmmmmmmm,-1,75,0,3,5

Prime95 v29.4.5.0 worktodo.txt shows:
PRP=xxxyyyzzz,1,2,mmmmmmmm,-1
PRP=aaabbbccc,1,2,mmmmmmmm,-1

I wonder if those last few bits are only required for double-checks when they need to match the first check.

The "75" is the factored-to-bits (or 99 if that's missing...which is true on some smaller exponents that have factors but were probably done with P-1 or ECM).

I don't know what that second one is... the zero.

"3" is the base... it'll probably always be 3 but I guess someone might have a program that does something different at some point.

And the final "1" or "5" is the residue type. That's important for double-checks because it needs to be the same residue type as the first check otherwise you won't be doing a valid check.

Maybe for first-time checks I should just leave those other parameters out and let Prime95 do its default thing?

Really though, I imagine in most cases you won't need the worktodo lines from this page at all. Only important if you managed to wipe out your txt file and need to add them back in. So... don't do that and you'll be fine no matter what.

[James Heinrich]

Originally Posted by Madpoo

"3" is the base... it'll probably always be 3 but I guess someone might have a program that does something different at some point.

Prime95 apparently supports any (int>=3) base for PRP, but defaults to 3 unless specified otherwise. Why you'd want a different base I have no idea, but it's there.

[Madpoo]

Originally Posted by Madpoo

...
And the final "1" or "5" is the residue type. That's important for double-checks because it needs to be the same residue type as the first check otherwise you won't be doing a valid check.
...

Took another look at past results...

PRP cofactor tests are coming in with residue types of 1 or 5, and non-cofactor tests are coming in with types 1 or 4.

Makes me think I should leave it out for this page (for first-time tests). Double-checks of course should still be set to match the previous test.

I ran a cofactor PRP test without those params and it used a residue type of 1 so that must be the default for those at least. Maybe a regular PRP uses 1 as well if not specified?

[Madpoo]

Originally Posted by Madpoo

...
I ran a cofactor PRP test without those params and it used a residue type of 1 so that must be the default for those at least. Maybe a regular PRP uses 1 as well if not specified?

I take it back. That test was with 29.3 and I realized I should be running 29.4 with the latest build 5. With that one it uses a default residue type of 5.

Which makes sense... that's (I think) the one that uses the Gerbicz error checking, and the residues can still be used to match previous tests when new factors are found. If I got that right.

Anyway, I'm just going to leave that residue type blank if there aren't any unverified tests waiting for a double-check. The default used will depend on the software in use.

[GP2]

Originally Posted by James Heinrich

Prime95 apparently supports any (int>=3) base for PRP, but defaults to 3 unless specified otherwise. Why you'd want a different base I have no idea, but it's there.

It might be useful to test PRP cofactors with other bases. I don't think there's been a case where a cofactor was 3-PRP but failed with some other base, but still.

For small exponents less than 64K, the cofactors have all been certified prime by Primo, so nothing further is needed.

For exponents larger than this but smaller than 500K, factordb has done PRP in various bases including at a minimum 3, 5, 7, 11. However, factordb does not test or record PRP-ness for cofactors of Mersenne exponents larger than this, for example M576551.

For exponents larger than 500K, paulunderwood has done some Lucas PRP tests (source code), including for the ones linked below:

• 576551
• 611999
• 675977
• 684127
• 696343
• 750151
• 822971
• 1010623
• 1168183
• 1304983
• 1629469
• 1790743
• 2327417
• 3464473
• 4187251
• 4834891
• 5240707
• 7080247

Also I imagine the Lifchitz PRP Top website probably does some independent testing before adding new entries, and other independent tests must have been carried out without formally recording them.

And of course for non-cofactor PRP testing, a subsequent LL test will certify primality, if and when we find one.

[GP2]

The Exponent Status report in text-only format is a bit garbled. It's missing a semicolon to separate the shift count from the residue type, and there is an extra semicolon at the end.

This is a recent change that seems to coincide with the switch from "CofactorPRP" to "PRPCofactor".

For instance, for M17509, here is the HTML format:

https://www.mersenne.org/report_expo...=17509&exp_hi=

The columns are: Status, Date, User, Residue, Number of known factors, Shift, Type, Base

Verified		2017-12-10	Oliver Kruse	F011B56D2685EC16	3	9334	1	5
Verified		2017-12-10	ATH		DF33C34A381E5D0E	3	16014	5	3
Verified		2017-12-10	ATH		F011B56D2685EC16	3	3196	1	5
Verified		2017-11-17	Oliver Kruse	DF33C34A381E5D0E	3	11413	5	3
Verified (Factored)	2017-11-17	Robin89		3076003BC0CD8C8F	1	0	1	3
Verified (Factored)	2017-11-08	kkmrkkblmbrbk	DF33C34A381E5D0E	2	12218	5	3
Verified (Factored)	2017-11-07	Oliver Kruse	DF33C34A381E5D0E	2	17291	5	3
Verified (Factored)	2017-10-12	kkmrkkblmbrbk	8255B2EBDE3838E0	1	16575	1	3
Verified (Factored)	2017-09-21	Oliver Kruse	8255B2EBDE3838E0	1	0	1	3

and the text-only format:

https://www.mersenne.org/report_expo...exp_hi=&text=1

The fields are: Exponent and record type and Status, Date, User, Residue, Number of known factors, Shift + Type concatenated, Base.

So for example, shift = 9334 and type = 1 becomes 93341

17509	PRPCofactor	Verified;2017-12-10;Oliver Kruse;F011B56D2685EC16;3;93341;5;
17509	PRPCofactor	Verified;2017-12-10;ATH;DF33C34A381E5D0E;3;160145;3;
17509	PRPCofactor	Verified;2017-12-10;ATH;F011B56D2685EC16;3;31961;5;
17509	PRPCofactor	Verified;2017-11-17;Oliver Kruse;DF33C34A381E5D0E;3;114135;3;
17509	PRPCofactor	Verified (Factored);2017-11-17;Robin89;3076003BC0CD8C8F;1;01;3;
17509	PRPCofactor	Verified (Factored);2017-11-08;kkmrkkblmbrbk;DF33C34A381E5D0E;2;122185;3;
17509	PRPCofactor	Verified (Factored);2017-11-07;Oliver Kruse;DF33C34A381E5D0E;2;172915;3;
17509	PRPCofactor	Verified (Factored);2017-10-12;kkmrkkblmbrbk;8255B2EBDE3838E0;1;165751;3;
17509	PRPCofactor	Verified (Factored);2017-09-21;Oliver Kruse;8255B2EBDE3838E0;1;01;3;