Thinking out loud about getting under 20M unfactored exponents

[Dylan14]

I received LaurV's request to stop work in the 11M range (I am that Delgado person).
Is 25.5 still available? If so, I would be happy to run it provided I am given appropriate bounds (unless it should be TF'ed further first).

[chalsall]

Originally Posted by Dylan14

I received LaurV's request to stop work in the 11M range (I am that Delgado person).

Thanks for checking in. It's easy to inadvertently "step on toes" when everyone's working "off-the-books".

Originally Posted by Dylan14

Is 25.5 still available? If so, I would be happy to run it provided I am given appropriate bounds (unless it should be TF'ed further first).

25.5M is not currently being "worked", but is scheduled for some additional TF'ing. No harm in running some P-1'ing on it now.

Another suggestion (if you don't mind the longer run-times) is some P-1'ing in the 3xM ranges. There are a bunch that only have had a B1==B2 run done on them, and they're all already TF'ed to 74.

Please see this report for some context, and then get some assignments by way of James' queries. Wayne is working in this area, but is currently doing deeper P-1'ing in the higher ranges.

[petrw1]

Yes 25.5 is available...however with SRBase scheduled for some serious TF there could be some toe stepping. You could wait a couple weeks until you see that it is at 73 bits.
At that time there won't be much P1 work required so the bound could be conservative.

30.5 is a good range if you can follow Yves. He is TF'ING to 75. So if you can P1 any at 75 bits you'd be safe.
Recommend bounds about 1M/30M.

Or freely P1 30.8 to about 1.5M/45M.

Thanks

Originally Posted by chalsall

Thanks for checking in. It's easy to inadvertently "step on toes" when everyone's working "off-the-books".



25.5M is not currently being "worked", but is scheduled for some additional TF'ing. No harm in running some P-1'ing on it now.

Another suggestion (if you don't mind the longer run-times) is some P-1'ing in the 3xM ranges. There are a bunch that only have had a B1==B2 run done on them, and they're all already TF'ed to 74.

Please see this report for some context, and then get some assignments by way of James' queries. Wayne is working in this area, but is currently doing deeper P-1'ing in the higher ranges.

[Viliam Furik]

My work on the 49.6M range, TF 74-75, is stopped for now. I may continue later, but for now, it should be considered free.

Sorry.

[SethTro]

Originally Posted by LaurV

We got unexpectedly clever after the 11.8M experience.
We moved to 11.4M, but we don't want to decrease the bounds - that would be a risky move and we may end up with not enough factors, like in 11.9M, therefore we kept the actual bounds and edited all the worktodo files in such a way to sort the 11.4xM sub-sub-ranges by the number of candidates. That is, we pushed the .42, .43, and .49 to the end, so if it happens that we find all the needed factors and we still have about 25% or 30% of the range unchecked, then the unchecked sub-sub-ranges will be those with less than 200 candidates left.

Clever huh?

So, the new world order is .48, .47, .46, .45, .44, .41, .40, .49, .42, .43 - just to avoid any toes. This will be done in 9 to 14 days, according to the resources available for P-1.

Trying to understand and extend this logic for myself. I'm working on 14.XX, some of the ranges (66 and 67) now have less than 200 factors.

As soon as a 14.XX range gets to 199 factors should I remove the remaining 14.XX range TF from my worktodo? or is that work still "valuable" for finishing the 14.X range before all the 14.XX ranges have finished?

[VBCurtis]

Seth-
If your range will easily make the 2000 goal, then I'd focus only on the subranges with more than 200 left. But if it's a tough range that will need a ton of P-1, I don't think it's worth e.g. taking TF to 74 on the 200+ ranges rather than just finishing the 2000 goal and moving to another big range.

Then again, the whole project is a bit of fun folly- so really the answer to your question is yours alone to decide. Once LaurV mentioned how he is going about things, I thought "hrmph, I should do that too".... but I'm using so little CPU on this that I haven't changed anything yet.

[LaurV]

Originally Posted by SethTro

As soon as a 14.XX range gets to 199 factors should I remove the remaining 14.XX range TF from my worktodo?

Nope, as Curtis pointed, you either let the rest of the sub-sub-range there, or move it to the end of the current sub-range, in such a way that if you find enough factors to get the sub-range under 2000, then you delete the rest of the sub-range (including what you moved at the end) and move to other subrange. Deleting them from worktodo before getting all the needed factors may be a mistake, you may need more work done and be forced to add them back in case you won't find enough factors.

In my case, I let untouched all the sub-sub-ranges which had over 200 candidates (and included), without bothering to sort them in a particular order (ideally they could be sorted descending, but they will need to be worked anyhow, most probably, and I didn't care) and I moved all the sub-sub-ranges with less than 200 candidates to the end, sorted in descending order of the number of candidates, as they were few, and not complicate to edit the worktodo. In this way, if I will find enough factors and I will stop any time, what will remain not-worked will be the sub-sub-ranges with less candidates. This may be a "small step" in a future (utopic) goal to get all the sub-sub-ranges under 200 candidates, but again, how you go about it is entirely up to you, and you don't need to go through all this trouble.

Right now, SRBase joining the fry with their supersonic factoring factory, screwed up all my order anyhow (but honestly, I am happy they are finding factors).

Some will say that the sub-sub-ranges with lower candidates have a higher factors density (they contained more factor in the past, that is how they became "less than 200 candidates", because more factors were eliminated), and excluding them (or pushing them to the end) may be a mistake, for the goal of "less than 2k factors in each subrange", but this is just a void argument, it may be true or not, but there is no proof one way or the other. Everybody does whatever runs his bike better...