ECM users - version 30.9-30.12 (see post#235) - Page 20

cmpute · 2023-03-09, 19:46

I can confirm this behavior. I just updated one of my computer from v30.8b17 to v30.10b4 and the automatically chosen bounds for P-1 tasks changed from B1=802000, B2=192756300 (on exponent 117256537) to B1=862000, B2=357135900 (on exponent 117257207). The program is set to use 84G RAM and the actual usage is similar before and after upgrading.

James Heinrich · 2023-03-10, 17:53

Quote:

Originally Posted by Prime95

-- PRP-CF done after new factor is found

Can you please confirm that PRP-CF will be done automatically, and before the factor is transmitted to PrimeNet? Same way that P-1 results are only sent after stage-2 is done, or ECM after all curves completed, the PRP-CF should be considered an extra stage of the factor-finding. There seems to be vocal support for this to avoid all possibility of poaching by those keeping a close eye on recently-returned factors.

Prime95 · 2023-03-10, 20:36

Quote:

Originally Posted by James Heinrich

Can you please confirm that PRP-CF will be done automatically, and before the factor is transmitted to PrimeNet? Same way that P-1 results are only sent after stage-2 is done, or ECM after all curves completed, the PRP-CF should be considered an extra stage of the factor-finding. There seems to be vocal support for this to avoid all possibility of poaching by those keeping a close eye on recently-returned factors.

The factor is sent immediately.
Then the PRP-CF takes place -- unlikely anyone will notice the factor and start a PRP-CF before you.
If ContinueECM is set, then rest of the ECM is then run.

chalsall · 2023-03-10, 21:16

Quote:

Originally Posted by Prime95

The factor is sent immediately. Then the PRP-CF takes place -- unlikely anyone will notice the factor and start a PRP-CF before you. If ContinueECM is set, then rest of the ECM is then run.

Extreme coolness. Thanks. There is a reason many have trusted George's code for many (many) years...

James Heinrich · 2023-03-10, 22:02

Quote:

Originally Posted by Prime95

unlikely anyone will notice the factor and start a PRP-CF before you.

But, depending how much effort they put in it, not impossible that the poacher finishes before the factor-finder.

Andrew Usher · 2023-03-11, 00:53

I think it is meant to apply to exponents small enough that that is extremely unlikely. George hasn't indicate yet what size exponent this is, but surely should make the limit configurable soon. The people that have a right to feel 'peached' are those who put in a lot more cycles than a PRP would take, which is usually ECM and the exponents practically always less than 5 million (such a PRP takes less than a day even on the slowest machines).

Andrew Usher · 2023-03-11, 23:37

peached -> poached

Maybe my first really bad typo here ...

(As for that exponent threshold, he has not yet updated the documentation at all, which is not unreasonable right now, and I could not find it in the source.)

seatsea · 2023-03-17, 10:47

My computer just hard restarted after running the latest version for a while, and this seems to have caused the prime.txt to be wiped clear. Thankfully did have a backup of the prime/local.txt files

James Heinrich · 2023-03-17, 11:44

Quote:

Originally Posted by Prime95

The factor is sent immediately. Then the PRP-CF takes place -- unlikely anyone will notice the factor and start a PRP-CF before you.

I've now seen a real-world example of this in action: 2800033

James Heinrich · 2023-03-20, 13:30

I noticed possibly-undesired behaviour from v30.11 doing auto-PRPCF on M41413. This exponent has 6 factors known since 2016 or earlier. JacobZ did a P-1, presumably(?) without specifying the list of known factors, and found a composite factor of the 4 smallest/oldest-known prime factors. Prime95 v30.11b1 then did an immediate PRPCF on those that 1 composite factor (which gets transmogrified to the 4 prime factors by PrimeNet on submission), but now the PRP-CF history section shows 4 tests with 6 factors, and then a more-recent test against only 4 of them. I assume it's partly the user's fault for doing P-1 with specifying known factors. But what if he had specified known factors and found a new one, would Prime95 do the PRP-CF on the list of known factors + new factors? Is it smart enough to check factor-found results against the list of known factors, divide out the known ones and be left with only previously-unknown factors (even, possibly, if the unknown portion might potentially a composite of 2+ new primes)? Is it possible for Prime95, when finding a factor, to submit the factor result (as it does already) and then ask PrimeNet for a proper PRP-CF assignment for that exponent, with the full list of all known factors (which will guaranteed be prime factors only, coming from primenet). If the machine is offline (transient or configured) then it can fall back to doing the auto-PRPCF based on what it knows, but it would be better if it had full data to work with.

Denial140 · 2023-03-20, 13:58

I have been doing a small amount of P-1 specifying previously known factors, and when a new factor is found the automatic PRP test runs with all specified and the new factor divided out; for example M300229.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
That's a Lot of Users!!!	jinydu	Lounge	9	2006-11-10 00:14
Beta version 24.6 - Athlon users wanted	Prime95	Software	139	2005-03-30 12:13
For Old Users	Citrix	Prime Sierpinski Project	15	2004-08-22 16:43
Opportunity! Retaining new users post-M40	GP2	Lounge	55	2003-11-21 21:08
AMD USERS	ET_	Lounge	3	2003-10-11 16:52

2023-03-09, 19:46	#210
cmpute "JZ" Mar 2023 37₁₀ Posts	I can confirm this behavior. I just updated one of my computer from v30.8b17 to v30.10b4 and the automatically chosen bounds for P-1 tasks changed from B1=802000, B2=192756300 (on exponent 117256537) to B1=862000, B2=357135900 (on exponent 117257207). The program is set to use 84G RAM and the actual usage is similar before and after upgrading.

2023-03-11, 00:53	#215
Andrew Usher Dec 2022 700₈ Posts	I think it is meant to apply to exponents small enough that that is extremely unlikely. George hasn't indicate yet what size exponent this is, but surely should make the limit configurable soon. The people that have a right to feel 'peached' are those who put in a lot more cycles than a PRP would take, which is usually ECM and the exponents practically always less than 5 million (such a PRP takes less than a day even on the slowest machines).

2023-03-11, 23:37	#216
Andrew Usher Dec 2022 1C0₁₆ Posts	peached -> poached Maybe my first really bad typo here ... (As for that exponent threshold, he has not yet updated the documentation at all, which is not unreasonable right now, and I could not find it in the source.)

2023-03-17, 10:47	#217
seatsea Jan 2021 2×3 Posts	My computer just hard restarted after running the latest version for a while, and this seems to have caused the prime.txt to be wiped clear. Thankfully did have a backup of the prime/local.txt files

2023-03-20, 13:30	#219
James Heinrich "James Heinrich" May 2004 ex-Northern Ontario 3×5×283 Posts	I noticed possibly-undesired behaviour from v30.11 doing auto-PRPCF on M41413. This exponent has 6 factors known since 2016 or earlier. JacobZ did a P-1, presumably(?) without specifying the list of known factors, and found a composite factor of the 4 smallest/oldest-known prime factors. Prime95 v30.11b1 then did an immediate PRPCF on those that 1 composite factor (which gets transmogrified to the 4 prime factors by PrimeNet on submission), but now the PRP-CF history section shows 4 tests with 6 factors, and then a more-recent test against only 4 of them. I assume it's partly the user's fault for doing P-1 with specifying known factors. But what if he had specified known factors and found a new one, would Prime95 do the PRP-CF on the list of known factors + new factors? Is it smart enough to check factor-found results against the list of known factors, divide out the known ones and be left with only previously-unknown factors (even, possibly, if the unknown portion might potentially a composite of 2+ new primes)? Is it possible for Prime95, when finding a factor, to submit the factor result (as it does already) and then ask PrimeNet for a proper PRP-CF assignment for that exponent, with the full list of all known factors (which will guaranteed be prime factors only, coming from primenet). If the machine is offline (transient or configured) then it can fall back to doing the auto-PRPCF based on what it knows, but it would be better if it had full data to work with.

2023-03-20, 13:58	#220
Denial140 Dec 2021 2⁴×5 Posts	I have been doing a small amount of P-1 specifying previously known factors, and when a new factor is found the automatic PRP test runs with all specified and the new factor divided out; for example M300229.