< 1M exponents
 

By the way, I got a bee in my bonnet about doing triple-checks on really small exponents that I thought it might be fun to do.

I tallied up all of the exponents below 1M that had only been double-checked. A lot of exponents in that range have had multiple checks because a) it's easy and b) sometimes people like to test small ones to get a quick run and verify the result against a known good value.

One of my reasons for doing so is that I wanted some small exponents to run on the other CPU of the system testing M383838383 and the smaller the better to keep contention down. The other was I thought it would be interesting to just check all of those as a "fun" verification of older work.

It was something like 11,200 exponents in total. It's down to something like 3000 left (all the ones above 830K or so).

Of course *after* the fact I realized that checking in so many in short periods of time means my results are going to flood the hourly stats of results on the server (the "recent results" report).

As a shout-out to anyone who might use that report (as opposed to the "recent cleared" which should be "madpoo free" since these have already been cleared), just send me a note and we'll work out something to get you the data you need.

I may spend some time today working on my grand plan of creating historical XML files that contain all of those results for each day, going back as far as there's data. That will solve issues like this where some dolt like me floods the server with results in a short period of time.

[QUOTE=Madpoo;399851]

(...) as opposed to the "recent cleared" which should be "madpoo free" since these have already been cleared)[/QUOTE]

For me, that´s really a non-issue, but just so you know, the "Recent Cleared Report" has not been "Madpoo free". Although the exponents were already DCed, they still show on the report. Thousands of them...

[QUOTE=lycorn;399854]For me, that´s really a non-issue, but just so you know, the "Recent Cleared Report" has not been "Madpoo free". Although the exponents were already DCed, they still show on the report. Thousands of them...[/QUOTE]

Oh... weird. I wouldn't have thought extra checks beyond double-checking would show up in there. That's a bummer.

Does the server accept LL results for exponents below 1M?

[QUOTE=Prime95;399865]Does the server accept LL results for exponents below 1M?[/QUOTE]

Well, it does because I told it to. LOL I know you have a limit in there to ignore LL results for exponents below 1M since they'd be kind of redundant. In some cases I can see why you did that... dozens of results for some of the small ones that people probably did as basic sanity checks.

I set it to cut off at 500K while I check in these triple-checks and then I'll set it back to 1M.

In most cases the triple-checks were entirely unnecessary, but in a select handful there were some where the only results might have been a pair where one of them was from one of those odd clients that sometimes had funky residues. In the cases I saw, the residue happened to match anyway (I'd looked for any that wouldn't have at the time) but now there's a solid pair of matching residues from clients that should have matching residues regardless.

I also saw a small number of cases where my own run of the exponent had an unexpected outcome. Basically, out of the 16 hex digit residue, the first 8 of those from my run were all zero, but the last 8 hexits matched the existing residues.

I can't really explain why that happened. I ran the test again for those small few (I think there were 5 out of the 11,100+) on a different system just to verify, and they came back okay on that next run.

So that was weird. Ever seen something like that before? I wonder if it was some odd side effect of having 20 threads running at once and it was more an issue with writing to the results file, or something else kind of strange happening. At least it was fairly consistent in that the first 8 hexits were all zero, so that was easy enough to spot.

[QUOTE=Madpoo;399870]I set it to cut off at 500K while I check in these triple-checks and then I'll set it back to 1M.
[/QUOTE]

And, all done. All exponents below 1M have been triple-checked at least. Nothing that new or interesting in the end.

[QUOTE=Madpoo;399871]And, all done. All exponents below 1M have been triple-checked at least. Nothing that new or interesting in the end.[/QUOTE]

If this means anything to anyone, just for later reference purposes...

I tested 11,180 (I think) exponents between 1000 - 1M. I had it running on several computers totalling 60 cores (3 servers, 20 cores each). Each test in it's own worker thread with just one core.

Total time was about 18-20 hours.

There are 61,415 exponents below 1M where a factor has been found, another 17,050 that have been LL tested, and of course 33 primes.

If we pretended that I was just going to LL test all 78,498 exponents, that means 60 cores running at 3GHz (E5-2690 v2) could have done LL on all exponents up to 1M in under 6 days. And that's not close to how many cores I could have used, it's just the ones that I had available at the time. :smile: Someone like Curtis with all of his systems could have whipped through even quicker. And that assumes I wouldn't even bother trying any TF at all, even the easiest stuff. Just doing that I would have found quite a few and saved a few seconds over doing a really fast LL test on the smallest ones.

I just thought it was thinking about how long it took GIMPS to get up to that milestone (on 1997-05-26 according to the milestone page).

Years from now maybe someone else will do the same look on the latest 1K core server (or more? Hello future Madpoo!) and could accomplish the same feat in under an hour.

Updated list
 

I ran off an updated list, attached here.

I'm also earmarking the 23M-24M range for myself.

[QUOTE=Madpoo;399870]I set it to cut off at 500K while I check in these triple-checks and then I'll set it back to 1M.[/QUOTE]

Seems unrelated but about the same time I notice my low TF assignments no longer show up?

They do NOT show as assigned to me...yet when the PC checks in (every day) it DOES NOT get the error that I do not own it and there is NOT a NA in my worktodo.txt.

In fact they have an assignment key....for example:

Factor=09408FAAA67E4BB6FEBC0B8EE251B4A4,89057,61,62

Or were they "Expired" when you checked in your Triple Checks?
Same date.

[url]http://www.mersenne.org/report_exponent/?exp_lo=89057&exp_hi=&full=1[/url]

That exponent was assigned on the 11/04/2015 to several users for ECM. That was the same date your assignment expired, and I think it is related (In those low ranges if an exponent is assigned for ECM work, any current TF assignement is "overriden"). I´m not 100% sure, but I would bet on it. Anyway, you may finish the TF assignment and the server will accept and credit the result.

[QUOTE=Madpoo;399473]

[snip]

< 20M are all up to 69 bits

[snip]

[/QUOTE]

You absolutely sure about that :no:

There's a load below 100k that are only at 61, I'm working through the low 3M's which are only at 65...