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

[Prime95]

30.8 build 15

Same as build 14 with a benchmarking bug fixed.

Windows 64-bit: https://mersenne.org/ftp_root/gimps/...8b15.win64.zip
Linux 64-bit: https://mersenne.org/ftp_root/gimps/...linux64.tar.gz
Source: https://mersenne.org/ftp_root/gimps/...b15.source.zip
Windows 32-bit: https://mersenne.org/ftp_root/gimps/...8b15.win32.zip
Linux 32-bit: https://mersenne.org/ftp_root/gimps/...linux32.tar.gz
Windows 64-bit service: https://mersenne.org/ftp_root/gimps/...64.service.zip
Windows 32-bit service: https://mersenne.org/ftp_root/gimps/...32.service.zip

[James Heinrich]

While experimenting with an exponent that doesn't really need P-1 factoring, I found some unexpected/undesirable behavior. I put in worktodo:
Pfactor=1,2,30759907,-1,67,4,"321626045933562402713,450961853042521358761,914209853451433227449"
Prime95 asked the server for AID, server said not needed, so Prime95 rewrote the assignment line with N/A which is fine, but in doing so it lost the list of known factors:
Pfactor=N/A,1,2,30759907,-1,67,4

Small related request, when an assignment line does include a list of known factors it would be nice if the worker window displayed the list of skipped factors during P-1 initialization:

Optimal P-1 factoring of M30759907 using up to 32768MB of memory.
Assuming no factors below 2^67 and 4 primality tests saved if a factor is found.
Already known factors: 321626045933562402713,450961853042521358761,914209853451433227449
Optimal bounds are B1=877000, B2=670821000
Chance of finding a factor is an estimated 12.1%

[Prime95]

Originally Posted by James Heinrich

While experimenting ... I put in worktodo:
Pfactor=1,2,30759907,-1,67,4,"321626045933562402713,450961853042521358761,914209853451433227449"

Pfactor= does not support a list of known factors.

[James Heinrich]

Originally Posted by Prime95

Pfactor= does not support a list of known factors.

Mea culpa.

Would it be hard to make Pfactor support known-factors list?

[James Heinrich]

Another obscure output request: If worktodo says Factor=<exponent>,72,73 the worker window says things like:

Starting trial factoring of <exponent> to 2^73
Trial factoring <exponent> to 2^73.
Trial factoring <exponent> to 2^73 is 0.090127% complete. Time: 11.789 sec.

It notably says the TF target level but does not mention the starting bit level. The output would be better with:

Starting trial factoring of <exponent> from 2^72 to 2^73
Trial factoring <exponent> from 2^72 to 2^73.
Trial factoring <exponent> from 2^72 to 2^73 is 0.090127% complete. Time: 11.789 sec.

I know Prime95 should rarely-to-never be used for TF these days, but should be a trivial fix.

[storm5510]

Originally Posted by Prime95

Pfactor= does not support a list of known factors.

But Pminus1 does, I believe.

Off-topic: gpuOwl does not support Pminus1. The best time I can get using Prime95 for a wavefront P-1 is around 8 hours. I can do it with gpuOwl in 2 using a 6.x version. I note the bounds Prime95 uses and set the gpuOwl bounds slightly higher.

[axn]

Originally Posted by James Heinrich

Would it be hard to make Pfactor support known-factors list?

Doesn't make sense. WIth PFactor, the bounds are determined by the number of tests saved. If it already has a factor, then why bother? Pminus1 has exactly what you need.

[James Heinrich]

Originally Posted by axn

Doesn't make sense. WIth PFactor, the bounds are determined by the number of tests saved. If it already has a factor, then why bother? Pminus1 has exactly what you need.

Sure it makes sense. It allows me to quickly and easily ask Prime95 to do a "good" or "great" or whatever kind of P-1 without me performing any kind of calculations as to what bounds are appropriate for that exponent. Pminus1 is good if you care exactly what bounds are used, but not very useful if you want Prime95 to auto-select the bounds (unless I'm missing something?).

[slandrum]

Originally Posted by James Heinrich

Sure it makes sense. It allows me to quickly and easily ask Prime95 to do a "good" or "great" or whatever kind of P-1 without me performing any kind of calculations as to what bounds are appropriate for that exponent. Pminus1 is good if you care exactly what bounds are used, but not very useful if you want Prime95 to auto-select the bounds (unless I'm missing something?).

On what basis is Prime95 supposed to decide what the bounds are? Tests saved will be 0 (there are already known factors), so this is obviously not meant to assist in the finding of Mersenne primes.

[James Heinrich]

Originally Posted by slandrum

On what basis is Prime95 supposed to decide what the bounds are?

On the basis of how many tests I tell Prime95 will be saved in the worktodo line. Same as always. Obviously no Mersenne primes will be discovered, nor candidates eliminated, when running P-1 on an exponent with known factor(s) but that's not at issue.

And you can't fault me for running P-1 on exponents that don't need it if George does the same thing.

[masser]

Originally Posted by slandrum

On what basis is Prime95 supposed to decide what the bounds are? Tests saved will be 0 (there are already known factors), so this is obviously not meant to assist in the finding of Mersenne primes.

One could make the argument that every additional factor found saves an eventual PRP check on the Mersenne cofactor.

To be clear, finding additional factors (for a specific exponent) does not help the Mersenne Prime search, but there are various other (much smaller) efforts that have some interest in the zoo of factors.