Factoring database

[Syd]

Originally Posted by henryzz

note that the time for 35 digits is also lower with a bound of 3M

Thats a good point. I've chosen the limit B1=1M only because it was higher than the "ECM high limits".
Now ECM very high limits starts with some P+1, P-1 and increases B1 up to 3M. After 2500 Curves with B1=3M it stops, on small numbers before that.

Are the limits reasonable now?

[smh]

Originally Posted by Syd

Are the limits reasonable now?

Yes, much better.

But since this starts with lower bounds, this makes people to push the "ECM very high limits" button straight away and thus kill any other users run, even if a small factor is found. Going manually through the different bounds often finds the factor and doesn't stop any other run.

[Andi47]

Originally Posted by smh

Yes, much better.

But since this starts with lower bounds, this makes people to push the "ECM very high limits" button straight away and thus kill any other users run, even if a small factor is found. Going manually through the different bounds often finds the factor and doesn't stop any other run.

I thought, when I click any button (no matter if it's "ECM low limits" or "ECM very high limits"), it will wait until a worker is available (i.e. until at least one of the other jobs is finished)??

BTW: Does the database keep track how much ECM, TF, P-1, etc. a composite has got? It would e.g. make no sense to do "TF to 1e5" twice, or to do 2000 curves at very low limits.

[Syd]

Originally Posted by Andi47

I thought, when I click any button (no matter if it's "ECM low limits" or "ECM very high limits"), it will wait until a worker is available (i.e. until at least one of the other jobs is finished)??

All jobs except the ECM very high limits have high priority, they are queued and processed once a worker has finished another job. If the worker is idle, it also picks up an ECM very high limits job and works on it until another job comes in. If lots of jobs are queued, ECM very high limits wont do anything.

Originally Posted by Andi47

BTW: Does the database keep track how much ECM, TF, P-1, etc. a composite has got? It would e.g. make no sense to do "TF to 1e5" twice, or to do 2000 curves at very low limits.

Yes it does. The TF buttons and the very high limits option disappears if it dosnt make sense. However, 2000 curves at very low limits are possible (if you take the time to click the button over and over again)

[mdettweiler]

Originally Posted by Syd

Yes it does. The TF buttons and the very high limits option disappears if it dosnt make sense. However, 2000 curves at very low limits are possible (if you take the time to click the button over and over again)

One thing I noticed about this: say, for example, you have a C100 that you run TF to 1e7 on. Of course, after that completes, the TF options will be shown on the work-done table, and the TF buttons will disappear...but, then, say you knock out a P10 factor with low-level ECM shortly after. Suddenly all of the work-done counts are erased, and TF options re-appear--even though, of course, the work done on the larger composite still counts just as fully for its cofactor.

I realize that this is primarily due to the way the DB tracks what work has been done on a given composite (since the new composite cofactor is tracked with a completely separate ID than the original number), but I would imagine it might nonetheless be possible to add some code so that, if a factor is found, any work done is automatically passed on to the respective DB entries of any composite cofactors remaining.

[mdettweiler]

Hmm...this is odd. I just tried running "ECM very high limits" on a C94 today, and noticed that it went until it finished the ~400 curves at B1=250000, and then stopped all of its own volition. I assumed at first that a new job had come in, and was about to restart it, but lo and behold--the ECM very high limits button had gone away! Is this by chance some new code that ensures ECM does not proceed past an optimal level?

[Syd]

Originally Posted by mdettweiler

Hmm...this is odd. I just tried running "ECM very high limits" on a C94 today, and noticed that it went until it finished the ~400 curves at B1=250000, and then stopped all of its own volition. I assumed at first that a new job had come in, and was about to restart it, but lo and behold--the ECM very high limits button had gone away! Is this by chance some new code that ensures ECM does not proceed past an optimal level?

Thats some new code I put in there, it should avoid wasting CPU time with too much ECM.

Something else: The results pages are now more compact, how do you like it? Changing it back would be just a click.

[mdettweiler]

Originally Posted by Syd

Something else: The results pages are now more compact, how do you like it? Changing it back would be just a click.

Hmm...not too bad. Though, there are some situations in which it would be better to have it turned off so you can see the full numbers...maybe a checkbox near the "Factorize" button?

[Andi47]

I just found this:

Number Work type Assigned 
9893068101...<84>  Sieve Assigned to worker #11 
2242206473...<80>  Sieve Assigned to worker #10

These two workers are offline - so why are there jobs assigned to them?

P.S.: Yesterday I noticed, that when a "ECM very high limits" job is running and I click the "ECM very high limits" button for another composite, the previous ECM very high limits job seems to be completely killed. I would suggest to either share the available workers to these low priority jobs, or put additional low priority jobs into a queue.

[smh]

Originally Posted by Andi47

I just found this:

Number Work type Assigned 
9893068101...<84>  Sieve Assigned to worker #11 
2242206473...<80>  Sieve Assigned to worker #10

These two workers are offline - so why are there jobs assigned to them?

I've seen that too, but the numbers do complete anyway.

Originally Posted by Andi47

P.S.: Yesterday I noticed, that when a "ECM very high limits" job is running and I click the "ECM very high limits" button for another composite, the previous ECM very high limits job seems to be completely killed.

Yes. And i noticed people going straight for the "very high limits". Often they find a small factor, but the other job is already killed by then. I've also seen composites smaller than 90 digits that have had the full 1200 curves at B1=1M. Maybe limit numbers < 90 digits to 500 curves or so?

[kar_bon]

what about this suggestion:

the very high ECM option will only be available, if some (has to be defined) low, medium, high ECM efforts are done!

for numbers upto about 80 digits it's quicker to use the sieve option instead of high or very high ECM!