Don´t step on my toes... please!

[lycorn]

I have recently reserved a bunch of exponents in the 929M range, for TF.
I checked to see whether they were reserved or not, and then registered them in the server (got AIDs for all of them). Today I was sending a batch of completed results, and got a fair amount of messages "Result not needed".
Ckecked the status of the exponents and verified they had just been tested by linded.
Could you please make sure the exponents you test are not reserved by anyone else? It saves time and resources...
Thank you.

[NBtarheel_33]

Originally Posted by lycorn

I have recently reserved a bunch of exponents in the 929M range, for TF.
I checked to see whether they were reserved or not, and then registered them in the server (got AIDs for all of them). Today I was sending a batch of completed results, and got a fair amount of messages "Result not needed".
Ckecked the status of the exponents and verified they had just been tested by linded.
Could you please make sure the exponents you test are not reserved by anyone else? It saves time and resources...
Thank you.

This happened to me this time last year with the 930M range. I think the entire 900M range is generally off-limits in the sense that linded (and perhaps lindee - they are father and son lumberjacks of the unfactored wilderness) has plans to work through it themselves ahead of PrimeNet. Monst has a similar thing going on @ 700M, and I think there is a concerted effort in the 500M range, too. AFAIK, anything left in the 200M range is open, as is the 300M range (below 332M), and the 400M range.

It's generally good GIMPSiquette to post in the LMH forum (when perhaps not feasible to actually register AIDs for every exponent) when you wish to work on/informally reserve large ranges. This helps to avoid crushed little piggies, and wasted CPU time. I agree that it can be quite frustrating to TF 10,000 exponents and have 80% of them rejected!

[Christenson]

One other point of GIMPSiquette: TF in the 3,320M range is handled by the OBD sub-forum.

[S34960zz]

If these other contributors wish to post their results to PrimeNet, then they should reserve the exponents prior to beginning their computations. Just like in baseball, "call the ball" so that another outfielder doesn't run into you. (900M+ is the outfield, eh?)

For fun and curiosity, I have taken on a handful (6 = 5 fingers and a thumb) of very large exponents, 5 in 900M+ and 1 in 600M+ territory. I'd be somewhat peeved to discover I'd burned 10+ days of GPU wall-clock time only to have the effort be of no use because someone else duplicated and submitted the work without having reserved the exponent.

I would have had no idea that linde[de] or anyone else had an unofficial claim to any ranges. Such claims should be coordinated with the PrimeNet mgmt. so that manual work reservations won't overlap.

[lindee]

Lycorn,

Sorry about stealing your work. About a year and half ago we got blocked for generating too much traffic when our cluster registered all its assignments immediately. Since then I have been pasting entire million-ranges into a queue where they don't get added into the worktodo files (and attempt to get assignments) until the nodes get short on assignments.

I haven't generally worried about situations like this one too much since most of the other LMHers have recognized us as having some sort of claim on the 900M range. I generally watch the TF reports on mersenne.info and skip over if I see any activity in a range we are coming up on.

I have not really looked into any way of mass reserving the exponents, but if anyone knows a way to do this without causing primenet any problems I would be glad to do it.

[NBtarheel_33]

Originally Posted by lindee

I haven't generally worried about situations like this one too much since most of the other LMHers have recognized us as having some sort of claim on the 900M range. I generally watch the TF reports on mersenne.info and skip over if I see any activity in a range we are coming up on.

I have not really looked into any way of mass reserving the exponents, but if anyone knows a way to do this without causing primenet any problems I would be glad to do it.

You're talking about hundreds of thousands of assignments (when I worked with the 930M range, there were in excess of 20,000 assignments, which blew PrimeNet's mind when I tried to register them all). I'd say the best way is to proceed as you have been doing, but perhaps we need a coordinated way (perhaps via a sticky in the LMH or Factoring forums) to make TF'ers aware of implicit claims like those of yours and Monst.

Are you aware of any mass-factoring-efforts outside of yours and Monst's that we might add to such a list?

[NBtarheel_33]

Originally Posted by S34960zz

If these other contributors wish to post their results to PrimeNet, then they should reserve the exponents prior to beginning their computations. Just like in baseball, "call the ball" so that another outfielder doesn't run into you. (900M+ is the outfield, eh?)

I liken GIMPS to probing the depths of space. At this point, 900M+ is "deep space", where the most we can do is catch a fuzzy glimpse of an object (i.e. TF an exponent to some level), rather than know intimately its properties (i.e. run an LL test and confirm primality). LL'ing 900M+ is basically akin to visiting Pluto. LL'ing an OBD candidate is leaving the Milky Way.

Originally Posted by S34960zz

For fun and curiosity, I have taken on a handful (6 = 5 fingers and a thumb)

Having lived in the Southern US, and hearing the requisite stereotypical jokes, I can't help but notice that your handfuls have 6 appendages. Y'all.

Originally Posted by S34960zz

of very large exponents, 5 in 900M+ and 1 in 600M+ territory. I'd be somewhat peeved to discover I'd burned 10+ days of GPU wall-clock time only to have the effort be of no use because someone else duplicated and submitted the work without having reserved the exponent.

Well, now that you're wise re: the Lindes' work in 900M, you might want to check with them and have them be aware of your exponents so that they might avoid testing them.

Originally Posted by S34960zz

I would have had no idea that linde[de] or anyone else had an unofficial claim to any ranges. Such claims should be coordinated with the PrimeNet mgmt. so that manual work reservations won't overlap.

I think they have discussed it before in the Factoring>LMH forum; that would be the only reason I would know about it. Also, if you check the "Recent Results" page, you will often see big players (e.g. Linded, Monst, etc.) turning in large quantities of TF results within a given range. This can be a good clue that there are implicit claims on certain ranges. I agree we probably need to have a big, bold sticky thread somewhere to coordinate this work, or perhaps George should just lock everyone out of certain ranges where known (and trusted) big players are working.

[NBtarheel_33]

Originally Posted by Christenson

One other point of GIMPSiquette: TF in the 3,320M range is handled by the OBD sub-forum.

I think the chance for toe-stepping is far less likely here, especially since the OBD ranges are well outside the GIMPS database limits, OBD work is not generating GIMPS credit (yet), and Prime95 cannot handle factoring OBD candidates.

[S34960zz]

handful (6 = 5 fingers and a thumb)

sort of like a baker's dozen, but for phalanges!

Originally Posted by NBtarheel_33

... or perhaps George should just lock everyone out of certain ranges where known (and trusted) big players are working.

Blocking out the ranges was what I had in mind but did not say exactly. Not sure whether it is easy or not-easy, for the PrimeNet administrators, to set a range as restricted so that exponents aren't offered from that range.

I've not been to the Factoring>LMH forum, though it sounds like I should at least visit briefly. I mostly watch PrimeNet, Math, Hardware>GPU, and Software within GIMPS; my schedule doesn't allow for much more than that. I count on the PrimeNet server to keep track of what work needs done.

I will drop a note to lindee on the 5-fingers'-worth of exponents I still have out. (Since I've already turned in two (finger M999999113 and thumb M667156621), I must have actually started with 6 fingers and that thumb!)

[LiquidNitrogen]

Originally Posted by lycorn

Today I was sending a batch of completed results, and got a fair amount of messages "Result not needed".

Couldn't you check the status before you start at this link:

http://mersenne.org/report_exponent/

[chalsall]

Lycorn et al... Further to lindee's post...

Those of us who do large amounts of independent LMH work (AKA lumberjacks; LOL) have learnt over the years how to avoid stepping on each other's toes. We tend to "take" 100M ranges, and work at a few 1M ranges at a time with that meta range, avoiding other's ranges.

Currently lindee "has" 900M, monst "has" 700M, and I "have" 500M.

Because we each do between 3000 to 7000 exponents per day, it would be impracticable to "officially" reserve them through the server. Both from the perspective of the work required at our end, but also the load it would place on the server.

Also, we tend to configure our clusters for several days or weeks of work, and then just let them go.

However, again, we try to avoid stepping on each other's toes, and when things got "tight" when the work was 63 to 64 "bits", we coordinated via threads in the LMH sub-forum reserving 1M or 10M ranges.

If you (or anyone) wants to work a range within the above three 100M ranges, I'm sure we'd welcome the help and would honour the request.

Keep in mind, as well, that very little LMH work is currently being done in the 300, 400, 600 and 800M ranges, so those would probably be pretty safe to attack. (Although someone did do ~3500 exponents in the 600M range over the weekend).

Perhaps we should again use the LMH sub-forum to "unofficially officially" reserve our ranges, even though things are still far from "tight"? (Knowing, of course, that no-one actually "owns" any range or exponent.)