(imo), I think 71 is fine, but hey.

In the winter time here, it can get [I]very [/I]crazy. Driving, you can hardly see ~25-50(approx) feet in front of you at times, quite dangerous.

Anyways, I am TF'ing to 71. :smile:

EDIT: No, I am not kidding.

[QUOTE=chalsall;355429]What isn't explicitly clear from the pictures is it continued to rain for the entire run and the "lime" afterwards -- Hashers, like GIMPers, can be strange people [U][B](see pic 80 for the three "hares", if you care)[/B][/U]....[/QUOTE]

So, who's who? Inquiring minds.....

[QUOTE=kladner;355442]So, who's who? Inquiring minds.....[/QUOTE]

LOL... There's a one in three chance I'm the one in the middle... :wink:

Inquiring minds don't just ask questions, but also do research...

The answer to this question can be determined with almost 100% certainty from publicly available information.... :smile:

Touché! :smile:

[QUOTE=chalsall;355429]Also, I didn't hear back as to what the best TF level is: 71 or 72? Right now there are [URL="https://www.gpu72.com/reports/available/lldc/"]28 candidates[/URL] TFed to at least 72 bits ready for assignment.[/QUOTE]
James' site is very accurate when it comes to my cards. With AMD cards (HD7970, HD7990), 72 is the number, as the LL test is a bit slower on those, you clean the exponents faster by doing TF to 72. For the amount of TF and P-1 done on those expos, you will probabilistically clear 2.2 exponents by TF-ing to 71, in the same time you do one DC with clLucas, and you will clear 1.1 exponents in the same time, when you TF to 72. Going to 73 is wasting time, you can only clean 0.5-0.6 exponents, therefore you will clean them faster doing DC. For a GTX card (580), where LL-ing is a bit faster comparing to TF, you will better stop at 71. You can clean about 1.3-1.5 exponents TF-ing to 71, and only 0.6-0.7 exponents when TF to 72, in the same time you do one DC. Anyhow, I TF-ed to 72.

(by cleaning one exponent, I understand either running a series of TF assignments, until you find one factor, therefore eliminating ONE exponent in all this time, or either getting a matching DC residue, therefore eliminating that exponent. I also included in the calculus the probability to get a DC mismatch, considered empirically to be about 1 in 15-20 tests).

[QUOTE=LaurV;355477]For a GTX card (580), where LL-ing is a bit faster comparing to TF, you will better stop at 71. You can clean about 1.3-1.5 exponents TF-ing to 71, and only 0.6-0.7 exponents when TF to 72, in the same time you do one DC. Anyhow, I TF-ed to 72.[/QUOTE]

OK, then from this, does that mean that you (and possibly others) would be willing to do the TF work to 72, even if those with cards like the 580 only do the work to 71?

What I'm trying to determine is at what level should I transfer them over to "Anon" for DCing assignment? 71, or 72?

Well, if we have to vote, I would vote for 71. I just finished an "unlucky bag" of 64 assignments (71 and 72 combined) from gpu72 and didn't find any factor, therefore losing a bit more than one day. Considering that those exponents had one LL done (so invested time for TF should be half), and also had P-1 done, I think releasing them for DC at 71 would be ok. Just my two PoE. (don't look for it, no way that google returns the right link :smile:, it is the smallest coin in PP game, you can search "yppedia").

[QUOTE=LaurV;355570]Well, if we have to vote, I would vote for 71. I just finished an "unlucky bag" of 64 assignments (71 and 72 combined) from gpu72 and didn't find any factor, therefore losing a bit more than one day.[/QUOTE]

OK, but remember than probability has no memory... :wink: Given an infinite (or even large) sample-set, this TFing work should have been profitable. It will be interesting to watch [URL="https://www.gpu72.com/reports/factoring_cost/"]the empirical cost per factor found[/URL] statistics for this sub-range...

As there are currently 34 candidates at 72 or above ready to be assigned for DCing, I'm going to hold off on transferring at 71 for a little while. I may do the "last bit" (read: to 72) myself to keep you guys fed as needed.

[QUOTE=davieddy;355571]WHY THE FUCK SHOULD YOU NEED TO ASK CHRIS?????

Who is in charge around here?[/QUOTE]

David-
He did not need to ask. He chose to. There is a cooperative community ethos in the GPU72 crowd. As mentioned in a previous post, Chris has some very utilitarian abilities which he shares with the group. There are no special privileges. Anyone could do these things. Chris has ready mechanisms for doing so with little effort. Since making a request to Chris is quick and easy, and he is willing to be accommodating, both the user and the project as a whole benefit.

I understand that there are differences of approach between you and Chris. I also understand that the both of you tend to be argumentative. However, argument does not need to involve invective and abuse. I'm sure that each of you has found the other irritating at times.

Nevertheless, when you go to all caps, and begin with "WHY THE FUCK....." it suggests deliberate provocation. I cannot address whether your posts are on full-time, or frequent advance moderation. However, a lack of [U]moderation[/U] in tone might bring about such a response.

You might also observe that there is a diminishing degree of tolerance for unnecessary turbulence among the regulars in these parts. You complain bitterly of being shut out, yet you don't seem to see that it is the waves you are making by thrashing and lashing about which are coming back to buffet you.

While there are other cases of animus on the forum as a whole, they don't generally infest the more technical areas. The GPU group, in particular, has a mostly cheerful tone, though there is competitiveness and teasing.

The Lounge is specifically grouped in Fun Stuff. When you persist in loud and abusive behavior, you are taking on the role, in pub terms, of a mean and obnoxious drunk. Like begets like. If you keep taking pokes at people, you can't really be too surprised if they eventually poke back. To carry the tavern metaphor one step further, you don't have real grounds for grievance should you find yourself on your ass in the street.

[CODE]And in the Euston tavern you screamed it was your shout
But they wouldn't give you service so you kicked the windows out
They took you out into the street and kicked you in the brains
So you walked back in through a bolted door and did it all again [/CODE]Sickbed of Cuchulain
-The Pogues
[YOUTUBE]h8ZAqxg__Rs[/YOUTUBE]

Can we get a report that has the wavefront (maybe 1 month and 3 month avgs) and how far we are ahead of it for each work type? No priority, I was just glancing around today to see where we were for p-1 and tf in the ll range and was thinking that it might be nice to see how we were doing for p-1, lltf, dc, and lmh in case someone felt like they needed to switch up their work.

[QUOTE=davieddy;355571]WHY THE FUCK SHOULD YOU NEED TO ASK CHRIS?????
Who is in charge around here?[/QUOTE]

Man, I can talk foul language better than you! I come from Romania :razz: we can go into Guinness book for the [URL="http://en.wikipedia.org/wiki/Romanian_profanity"]language with the most profanities[/URL] reported to the total number of words, and be aware that Latin-derived languages have more words than other languages, due to the flexible word terminations (we have about one million words in the scrabble-accepted word list, which include all words with all their forms, contrary to English with has only about 268 thousand words in the [URL="http://en.wikipedia.org/wiki/SOWPODS"]SOWPODS[/URL]). (I played Scrabble "pro level" 30 years ago, when SOWPODS only had 150k words, and its Ro version only 640k).

For sure George is in charge, but fuck me if I would bother him to ask for assignments. I would better know him doing improvements on P95 than looking for whatever numbers to give me to cook. On the other side, Chris has everything automated and he is very cooperative. My choice to do 37M is MY choice, and I don't give a fuck on what you, Chris or George says about. They can help me if they want, or not. I chosen the path of the least resistance, to see my goal done. I could do it on private, but I chose to do it publicly because I think the advice (of clLucas and 37M) is useful for other people too.

My big boss own my company, but I don't go to him for any freaking piece of paper, pencils, company car's keys, etc, every time I need such things. Wake up David! And come to your senses. Or are you envious that I didn't ask you first?

[edit 2: on the other hand, you should be happy, because me voting for 71 instead of 72 comes in line with your theory, and not with Chris' one :razz: But I always say what I think, regardless of who else is arguing on the subject]

[QUOTE=Aramis Wyler;355693]Can we get a report that has the wavefront (maybe 1 month and 3 month avgs) and how far we are ahead of it for each work type?[/QUOTE]

Interesting thought... :smile:

I have something crude which might slightly help present this... Compare [URL="http://www.mersenne.info/exponent_status_line_graph/1/0/"]this graph of exponent status[/URL] with [URL="http://www.mersenne.info/trial_factored_min_max/1/0/"]this one showing factoring depth.[/URL]

Unfortuantely I don't show P-1 data on Mersenne.info. Need to add that to my ToDo list (They just never get shorter! :cry:).

It would be interesting combining all of this data (plus additional, like assignment status) into a single graph / table / report.

(James: want to beat me to it? I wouldn't complain... :wink:)

Edit: Forgot to mention... Yes, I know about the error in dataset in the ~33M to 34M range... Mersenne.info needs some TLC....

Really thinking that we need some moderation/rules of conduct w.r.t. language on the forum. This project is something that is supposed to be worldwide, open to all ages, and all sensibilities, in its welcome to participate. Moreover, George has placed explicit faith in this forum and its users by including the forum URL as an "official" source for help in the Prime95 help menu. I believe I may have raised this point before, but if the first thread that a newcomer encounters is filled with vitriol and foul language, they are very likely to dismiss the whole project as being a joke, lost to Eternal September. If the newcomer is a kid and their parents should happen to witness some of this not-so-witty banter, we've lost a whole family's respect for GIMPS - and the way mothers gossip - likely a whole school district's respect for GIMPS.

Are we a serious mathematical distributed computing project, or an entry on the NetNanny black list? Didn't screaming swears on the Internet become passe around 1997?

David, you don't like GPU72. You might not even like Chris all that much. We get it. So stay out of the GPU forums (or perhaps the mods can *keep* you out), and post in other areas of the forum where you can offer something positive.

GPU72 has been adopted by users and tacitly endorsed by George. It is (for now) a law of the land. Deal with it. (Hmm, sounds eerily similar to the US government...)

I really think some of the mods here need to get that rarely used hammer out of their closet and use it....

[QUOTE=kracker;355760]I really think some of the mods here need to get that rarely used hammer out of their closet and use it....[/QUOTE]

Reliable sources inform me that he is gone for six months.

[QUOTE=Aramis Wyler;355693]Can we get a report that has the wavefront (maybe 1 month and 3 month avgs) and how far we are ahead of it for each work type? No priority, I was just glancing around today to see where we were for p-1 and tf in the ll range and was thinking that it might be nice to see how we were doing for p-1, lltf, dc, and lmh in case someone felt like they needed to switch up their work.[/QUOTE]

[QUOTE=chalsall;355753]Interesting thought... :smile:

I have something crude which might slightly help present this... Compare [URL="http://www.mersenne.info/exponent_status_line_graph/1/0/"]this graph of exponent status[/URL] with [URL="http://www.mersenne.info/trial_factored_min_max/1/0/"]this one showing factoring depth.[/URL]

Unfortuantely I don't show P-1 data on Mersenne.info. Need to add that to my ToDo list (They just never get shorter! :cry:).

It would be interesting combining all of this data (plus additional, like assignment status) into a single graph / table / report.

(James: want to beat me to it? I wouldn't complain... :wink:)

Edit: Forgot to mention... Yes, I know about the error in dataset in the ~33M to 34M range... Mersenne.info needs some TLC....[/QUOTE]
Thinking about one aspect of progress, I looked around to see if there was a table of factored depth by range just for no-LL exponents, but had no luck. (Yeah, I know, it's probably in there somewhere, but I just couldn't find it.) Could we see the numbers from Primenet that are behind the [URL="http://www.gpu72.com/reports/estimated_completion/primenet/"]estimated completion[/URL] table perhaps? P-1 / no p-1 would be nice too. Anyway, something from my wish list to weigh up for your to do list.

[QUOTE=garo;355869]Reliable sources inform me that he is gone for six months.[/QUOTE]

I would say that a *permanent* vacation from the GPU subforum might be in order, as well.

[QUOTE=markr;356012]Thinking about one aspect of progress, I looked around to see if there was a table of factored depth by range just for no-LL exponents, but had no luck. (Yeah, I know, it's probably in there somewhere, but I just couldn't find it.)[/QUOTE]

That could be indirectly derived from the reports from Mersenne.info, but not directly -- it would take some work.

[QUOTE=markr;356012]Could we see the numbers from Primenet that are behind the [URL="http://www.gpu72.com/reports/estimated_completion/primenet/"]estimated completion[/URL] table perhaps? P-1 / no p-1 would be nice too. Anyway, something from my wish list to weigh up for your to do list.[/QUOTE]

An excellent suggestion, and something quite easy for me to build a query for (all the needed data already exists in the GPU72 MySQL database, updated in real-time). Please consider this on my "Short-list" ToDo list!

[QUOTE=NBtarheel_33;356030]I would say that a *permanent* vacation from the GPU subforum might be in order, as well.[/QUOTE]

If six months of reflection is insufficient to change behaviour, a permanent vacation may be in order.

[QUOTE=chalsall;356038]An excellent suggestion, and something quite easy for me to build a query for (all the needed data already exists in the GPU72 MySQL database, updated in real-time). Please consider this on my "Short-list" ToDo list![/QUOTE]

OK, the first draft of the requested [URL="https://www.gpu72.com/reports/current_level/"]Current Trial Factoring Depth for all Candidates[/URL] report is now online.

The table for LL candidates is complete. The DC table below it needs some attention, but I figured the LL data was what most people were interested in, so I figured I'd release the report now.

The back-end is also calculating the data for those candidates without P-1 done. I'll expose that once I get the DC table finished (it will be available via a sub-menu for the new "Current TF Levels" menu).

Coolness! Thanks! :tu:

[QUOTE=chalsall;356135]OK, the first draft of the requested [URL="https://www.gpu72.com/reports/current_level/"]Current Trial Factoring Depth for all Candidates[/URL] report is now online.[/QUOTE]The 33M-47M (DC?) table shows counts under the bit depth and total columns as '%.4f' rather than thousands-separated integers like everywhere else in the table.

[QUOTE=James Heinrich;356147]The 33M-47M (DC?) table shows counts under the bit depth and total columns as '%.4f' rather than thousands-separated integers like everywhere else in the table.[/QUOTE]

Yes... That's the "The DC table below it needs some attention..." reference above...

For expediency I just re-used the code which renders the "Estimated Completion" report. I haven't yet converted the DC table to render the data as INTs rather than FLOATS. Later today.

The whole verbal abuse and fighting on a personal level is the reason I have opted out of some of the discussion topics. The issue of a younger child coming in and having him and his mother see a bunch of supposed adults swearing at each other is one thing. Sure, I'll admit to having used one slice of bad language fairly recently on this forum (in fact it spawned from the same hopeless discussion as most of the recent bile) but it was of course meant for the same purpose as any other word: to convey some kind of message. In this case, the message required special emphasis that only comes from words which are meant to be used less frequently, lest they lose that emphasis.

One still has to be careful because taken out of context, a few posts of "emphasized wording" can make the place look seriously unkempt. Sadly I think the last couple of days have seen the F Word used a dozen times. We're getting to that stage where it is becoming impossible to justify any of it.

Worse still I think is some of the attitude behind the swearing. As someone mentioned, this IS an official source for help on all things GIMPS (and more, yet) and people coming here are going to be impressionable on account of they [I]want[/I] answers. They're extra sensitive to information, just as I once was just a few short months ago. Since then, I've been absorbed into the Hell-spawned TF depth debacle, and I'm quite lucky to have been drawn in by two conflicting posts with fairly representative levels of information from either side. The angry old guy isn't making much sense, and the friendly old guy is presenting data. I was fortunate enough to be able to choose the "right" side before things got nasty.

On the other hand, when someone who just doesn't know when to quit keeps going on about something while everyone who is nearby happens to not give the two flying, uh, ducks, to say "you're wrong and here's why" for the seventeen bollillionth time, you're going to get people choosing sides. I almost felt sympathy for David when he pulled the "I'm being bullied and mocked and still I stand true to my cause card" until I realized he'd been doing this to himself for such a long time. If there was any left, it's all gone now, since he's now resorted to directly coarse language.


The environment has turned completely caustic and I feel bad for anyone entering now, expecting something very different. I'll do my best to keep the noise down and the signals clear, but there's only so much one can do.

@ TheMawn- The situation is regrettable, and I do feel bad for David, but not to the extent of tolerating the destructiveness which was being stirred up. I'm glad you stuck it out. I would also like to think that with the provocation put in Quarantine, that the waves will settle down, and people can have reasonable discussions, including disagreements, without going for the jugular.

It is truly a shame that David is too much in turmoil for polite company. To be blunt, I think mortality is heavy on his mind. He mentioned liver failure some time back. If he's trying to control a long-standing alcohol habit, that can bring about all kinds of emotional backlash.

I don't mean any of this to excuse bad behavior, but only to offer my understanding and speculation about the background of that behavior.

Thanks TheMawn, klander et al...

I must apologize for being a bit of a lightening rod for some of the negative energy here on the forum.

I truly do not understand David's motivation. But, to add to klander's thoughts, it is possible that David considers the idea of GPUing to 72 to be his and he thinks I "stole it" from him. I really don't remember who within the community initially suggested that particular number -- it might have actually been David.

However, for the record, I called it GPU to 72 (and registered the domain) because it rhymed. Further, I would argue that the GPU72 system was inspired by the fact that Mr. P-1 was initially coordinating the work by hand with a dozen or so people with GPUs.

I believe strongly that argument (read: debate) is healthy (and, you might have noticed, I really enjoy it). One might just learn things. However, debate (read: argument) should be respectful, rational and sincere.

There is a non-zero probability that one or the other side(s) are wrong.

There are situations where there is a non-zero probability that everyone is wrong.

Namaste to all.

[QUOTE=LaurV;355232]Hey Chris, can you pull in some 37M-38M exponents for LLDC?...[/QUOTE]
This was the post that was the nail in the coffin. What I don't understand is why [I]he[/I] got so upset? If you want a change to the way G72 works, [I]who else [/I]would you ask to make the change? G72 workers were asking for changes to G72 assignments, period. Though I also enjoy the usual back and forth, it did get out of hand.

Look forward to future, healthy debates :smile:

I wholeheartedly agree to that. I just hate it when it gets "out-of-hand".

Chris- It is one thing to suggest some form of an idea. It is quite another to implement something which accomplishes the aim. It is yet another level to take on a never-shrinking To Do list in the development and polishing of that implementation. If David has the programming chops, he could have put code where his "mouth" has been.

When I got my first GPU, I put it to work in Mr P-1's group. Reading about the wonders of GPU factoring probably had a lot to do with getting the card. Besides, I got a close-out deal on the GTX 460 which I did not even try to refuse.

It amazes me what Mr P-1 was doing, apparently by hand, in capturing newly-released exponents, and farming them out to workers via email.

I certainly remember when you (Chris) started talking about your spider. I was rather too ignorant to attempt joining in at that point. I think I jumped in almost as soon as you put up the earliest GPU to 72 pages.

Just for those who came later, or don't remember: the last I knew, Chris also provides the server space for everything associated with GPU72. He sort of laughingly started a Team which now sits among the highest rankings in a number of production categories.

No doubt, people with the hardware would have found ways to keep it fed. It seems likely to me that there would not be as many of them, and that it would be a lot more work for them to keep running.

When Scott (swl551) made the beginnings of MISFIT available, that really completed the system we have now. He has put an awful lot into fulfilling people's wish lists, too. I happen to be one of the people who, as Scott put it, "Likes to fondle his exponents," so I don't use nearly all the automation he built. It really has become a "set and forget" utility. Even so, it is a super-convenient way to handle things manually, too.

So, I'm just saying that I see this whole sub-project as a shining example of the best that GIMPS has to offer. There are many others; more than I can name off the top of my head, so I hope those I don't remember will forgive me and my aging brain cells. Others should feel free to add to this honor roll.

Of course, George and Scott K. got things going, and head the pantheon. Just thinking about the GPU end of things I come up with
msft
The Judger
Bdot
chalsall
kracker
dubslow
frmky
jasonp
henryzz
LaurV
flashjh
firejuggler
nucleon
TeknoHog
xyzzy (though he might not admit it, this sub-forum is proof.)
Brain
axn
garo
xilman
bcp19
ET
ATH

I consider that people who engaged in discussion, suggested features, beta tested, etc., all contributed to the overall effort. Don't take the order of the list as meaningful except at the very top, where even I have some idea of who the real titans are.

You forgot owftheevil, creator of P-1 for CUDA. :smile:

TheJudger=mfaktc
Bdot=mfakto
msft=CUDALucas(Dubslow as well here), clLucas

EDIT: Ugh for all things why am I in that list??


[SIZE=1] (Also, where's daviddy /rofl)[/SIZE]

[QUOTE=kracker;356224]You forgot owftheevil, creator of P-1 for CUDA. :smile:

TheJudger=mfaktc
Bdot=mfakto
msft=CUDALucas(Dubslow as well here), clLucas

EDIT: Ugh for all things why am I in that list??


[SIZE=1] (Also, where's daviddy /rofl)[/SIZE][/QUOTE]

I knew I would forget someone. Thanks, kracker. As to your presence, you have been a big part of testing owtheevil's work.

I don't know how to parse David's contributions. Insufficient data on my part.

[QUOTE=kladner;356236]I knew I would forget someone. Thanks, kracker. As to your presence, you have been a big part of testing owtheevil's work.

I don't know how to parse David's contributions. Insufficient data on my part.[/QUOTE]

:ermm: I haven't tested/done anything here/there, I only have OpenCL cards.(at the moment)

I guess my mention was based on the impression that you've been quite involved in the discussions. No embarrassing implications intended. :smile::blush:

[QUOTE=kracker;356224]EDIT: Ugh for all things why am I in that list??
[/QUOTE]
Same here too. Beside of having a ton of hardware and a big mouth, I didn't contribute too much... Beside of the titans (like msft, Oliver, Bdot, oftheevil, who contributed with nice software) I would stress more on Jerry (flash) and Bill (Dubslow) who created the "modern version" of cudaLucas, and to James (hey, where is he in the list?) who did the "mersennaries" site (actual mersenne.ca). Also, where is Scott-misfit? I would also put jasonp, bb, and few others in a top of a separate list, less related to mersenne/gimps and more related to factoring in general, and I would make a third list with people who really invested "hardware time", like Craig (nucleon), Mike (Xyzzy), monst, and few others. On the third list, somewhere in the middle, I would place myself. :smile:

I told you I would overlook some people. James has crossed my mind more than once while I have been away from the computer. I sincerely regret the neglect.

You LaurV, on the other hand made my list [I]because[/I] of that big mouth. :razz: (The quieter ones are easier to slip up on.) That, and your detailed descriptions of tuning FFT lengths, and other user-friendly information stuck in my mind. Take it as you will. I think I saw a good bit of contribution to discussion, feature concepts, and general good humor.

[QUOTE=kracker;356238]:ermm: I haven't tested/done anything here/there, I only have OpenCL cards.(at the moment)[/QUOTE]

OK. Maybe my confusion had to do with the area of involvement. I see a lot of feedback from you in Trial Division with OpenCL (mfakto).

[QUOTE=chalsall;356135]OK, the first draft of the requested [URL="https://www.gpu72.com/reports/current_level/"]Current Trial Factoring Depth for all Candidates[/URL] report is now online.

The table for LL candidates is complete. The DC table below it needs some attention, but I figured the LL data was what most people were interested in, so I figured I'd release the report now.

The back-end is also calculating the data for those candidates without P-1 done. I'll expose that once I get the DC table finished (it will be available via a sub-menu for the new "Current TF Levels" menu).[/QUOTE]
Lovely! That shows the picture well - thank you.
And that was done quickly! (I was away from the forum for a few days, so now I'm replying to posts two pages back!)
[QUOTE=chalsall;356151]For expediency I just re-used the code which renders the "Estimated Completion" report.[/QUOTE]
I hoped that report would be a good place to start from. Thanks again.

Pls let me replace "Manual" on the "Get Assignments" form
 

Chris, please allow me to request (again) some feature that allows me to replace the "Manual" computer name when requesting work. I now have (at least) 50 assignments that were probably lost in network timeouts. Knowing for which machine they were requested would save me a lot of trouble determining if they really were lost.

Unfortunately I cannot use any of the automation tools for most of my machines ... therefore almost all my assignments are "Manual" and I have to search all machines for traces of the aging assignments :-(.

[QUOTE=Bdot;366209]Unfortunately I cannot use any of the automation tools for most of my machines ... therefore almost all my assignments are "Manual" and I have to search all machines for traces of the aging assignments :-(.[/QUOTE]

Copy you. Thanks for the "squeak". :smile:

I'll try to get to this this weekend.

www.mersenne.info often out of sync
 

I sometimes see that statistics are one, two or more days out of date (beyond the apparent 24 hour update cycle), over many months now.
Further, I note that on such occasions, looking at other regions of the exploration space, I see that the information can have a different level of currency, so the evidence I see is that such outages are not across the whole space.
This would be consistent with an interrogation of the PrimeNet server to scrape statistics, but upon communication delays such process doesn't entirely recover to adequately retry and fulfill regions.
Please don't tell me that the current mechanism is a http viewable page scrape!

With increasing reliance on .info, would it be useful to negotiate a more efficient and reliable issue of information from PrimeNet?

[QUOTE=snme2pm1;376291]I sometimes see that statistics are one, two or more days out of date (beyond the apparent 24 hour update cycle), over many months now.[/QUOTE]

Could you please give specific examples? URLs welcome.

Please note that mersenne.info goes out of its way not to negatively impact mersenne.org, and thus does not spider those ranges which have not had a state change (as previously documented).

[QUOTE=chalsall;376327]Could you please give specific examples? URLs welcome.

Please note that mersenne.info goes out of its way not to negatively impact mersenne.org, and thus does not spider those ranges which have not had a state change (as previously documented).[/QUOTE]

Well, I had best look at this previously documented algorithm that is alluded to before further comment.
A link to your algorithm description would help.

Regarding recent observations, which I wasn't in a hurry to particularly emphasise, you might recall that I have been working the early 78M region, and almost completed such thrust.
Today again [url]http://www.mersenne.info/trial_factored_tabular_delta/4/78000000/[/url]
is not aware of changes, though I certainly did lodge some results that I imagined would have been noticed.
Back a few days, I had the impression, without recording precise information, that such was three days without update.

[QUOTE=snme2pm1;376335]Today again [url]http://www.mersenne.info/trial_factored_tabular_delta/4/78000000/[/url]
is not aware of changes, though I certainly did lodge some results that I imagined would have been noticed.[/QUOTE]

The spider will only "notice" work (in a 1M range) if there's a state-change on [URL="http://mersenne.org/primenet/"]the Primenet Exponent Status Distribution[/URL] report.

A state-change is defined as a change in the number of factors, the number of LL/DC candidates, or the number of assignments (of any class -- TF/P-1/LL/DC). Thus, if you didn't actually get the assignments from Primenet AND didn't find a factor, the system won't detect any TFing changes.

Note that the system also scans 30 additional 1M ranges without a state-change, sorted by last scanned date. Thus, a range which has had TFing work done on it without a factor being found OR an official assignment being made, it can take up to ~30 days for the range to be updated.

Thus, if you want a range to be scanned simply get an assignment from Primenet and hold onto it for 24 hours.

I notice that the /account/factoring_cost/ only shows exponents up to 70M. It would be nice to have it show up to 74M to include the current TF assignments being handed out. :)

I would be weighting in favor of adding few lines to the [URL="http://www.gpu72.com/reports/current_level/"]current levels[/URL] DC table too... (like to 55M or so? is there any work to be done there, or it was already stomped all when we did LLTF in that range?)

[QUOTE=Mark Rose;379042]I notice that the /account/factoring_cost/ only shows exponents up to 70M. It would be nice to have it show up to 74M to include the current TF assignments being handed out. :)[/QUOTE]

Thanks for pointing that out. Fixed.

[QUOTE=LaurV;379092]I would be weighting in favor of adding few lines to the [URL="http://www.gpu72.com/reports/current_level/"]current levels[/URL] DC table too... (like to 55M or so? is there any work to be done there, or it was already stomped all when we did LLTF in that range?)[/QUOTE]

I've expanded the "view" to go to 60M, on this report as well as the "Estimated Completion" reports.

To answer your question, there is still DCTF'ing work to be done above 50M, although mostly below 55M. Also, this will probably have to be adjusted (to deeper TF'ing) once DC'ing gets into the 50M range (in a few years).

Lastly, please note that the numbers in these reports will not correlate entirely with Primenet's report of Exponent status. Because of the queries involved (which are quite expensive (for GPU72, not Primenet)) any candidates above the current "release level" are not currently included in the totals.

Thanks!

Well, indeed, after some calculus and few tests, the 57M DCTF and higher, have to go to 73 bits, i.e. the last 3 squares in the "72" column have to be white. And if/when you offer those for grabbing, I will take some, for my 7970 (which is slow for 74 bits, but quite ok for 73 and lower).

Would it be possible to create a page like [url]http://www.gpu72.com/reports/workers/day/[/url] that shows yesterday's work done? That way it would be easy to compare a whole day. :)

[QUOTE=Mark Rose;380246]Would it be possible to create a page like [url]http://www.gpu72.com/reports/workers/day/[/url] that shows yesterday's work done?[/QUOTE]

Yes.

If I can help in any way, please let me know.

Would it be possible for Primenet to upload a .txt file instead of displaying the data in the web page, wherever the option exists for "display simple text format only" or whatever the exact words are?

With the server seeing a big upgrade (and working very nicely with the workaround thus far!) I'm wondering if the 50,000 lines of text limit could be raised. It would be even more cumbersome to actually use on the web page but the .txt file could have 500,000 lines and be easy to use.

[QUOTE=TheMawn;380710]Would it be possible for Primenet to upload a .txt file instead of displaying the data in the web page, wherever the option exists for "display simple text format only" or whatever the exact words are?

With the server seeing a big upgrade (and working very nicely with the workaround thus far!) I'm wondering if the 50,000 lines of text limit could be raised. It would be even more cumbersome to actually use on the web page but the .txt file could have 500,000 lines and be easy to use.[/QUOTE]

It think it's set that way for two reasons:

1. To avoid the default POST request maximum size in PHP/IIS. This could be easily changed.
2. To avoid running into a PHP timeout processing all the results. This might not be so much of an issue with the new server.

Not sure why this is in the GPU72 feature request thread though ;)

[QUOTE=TheMawn;380710]Would it be possible for Primenet to upload a .txt file instead of displaying the data in the web page, wherever the option exists for "display simple text format only" or whatever the exact words are?[/QUOTE]Are you referring to submitting results to PrimeNet, or getting some kind of report data from PrimeNet?
Result-submission will be reworked will be completely reworked in the near future when the hardware migration is complete. If you were talking about reports, then an example of what you mean would be useful. Although perhaps not in this thread since this is for GPU72. :smile:

Oops. I didn't notice that this feature request thread was in the GPU72 section.

Hmm, How about a 3rd option in Misfit of "LetGPU72Decide" instead of First Time / Double Check TF...

[QUOTE=NickOfTime;381051]Hmm, How about a 3rd option in Misfit of "LetGPU72Decide" instead of First Time / Double Check TF...[/QUOTE]

A good idea... Although because of the evolutionary nature of GPU72, the APIs for the assignments use distinct URLs for the two worktypes -- DCTF was a secondary thought, implemented by request of participants.

To put on the table though, I already do this with LaurV and Jerry (with their knowledge and permission) -- the LLTF form will assign DCTF iff needed. If you (and/or anyone else) is willing, I can in the short term implement this individually. Please PM me with authorization to send your clients DCTF when needed (this will only be done when LLTF LGD is set as the option, and not very often).

Longer term, if Scott is willing to implement this in MISFIT, I could add a new worktype to the LLTF form (say, option 10) which (like option 9: LGD) is a sort of "Super LGD" which chooses the WorkType in addition to the Range and Depth.

Thanks for this suggestion. Scott: thoughts?

As I understood, if you use "Let GPU72 decide", then GPU72 can decide! This means that you can get either LLTF or DCTF assignments, regardless what you have selected in that box. If you want to get your specific assignment (like either LLTF or DCTF) then use "What Makes Sense". For even "more specific", use the other options. Misfit already specifies the fact that the server can change your assignment depth, but it is a bit confuse because it doesn't gray the "Type" box too, when "Let..." is selected instead of "WMS".

To be more specific a third option for "type" would also be ok. Like "I don't mind what you give me to do" hehe..

To avoid any confusion, all my cuda cards are on "Let GPU72 decide" and I understand by that the fact that GPU72 can give me to do whatever it considers necessary, bitlevel, range, [U]and type[/U], but the opencl cards are on 73 bits max, because mfakto 74++ bit kernels are lousy (i.e. much slower compared with lower-bits kernels).

[QUOTE=LaurV;381057]As I understood, if you use "Let GPU72 decide", then GPU72 can decide! This means that you can get either LLTF or DCTF assignments, regardless what you have selected in that box. If you want to get your specific assignment (like either LLTF or DCTF) then use "What Makes Sense". For even "more specific", use the other options. Misfit already specifies the fact that the server can change your assignment depth, but it is a bit confuse because it doesn't gray the "Type" box too, when "Let..." is selected instead of "WMS". [/QUOTE]

OK. Good point.

I perhaps overly constrained myself (and ourselves) -- I ASSuMEed that those who choose LLTF want to only do LLTF, and those who choose DCTF want to only do DCTF. I know this was true for some for a while (Pete: we still miss you!!! :sad:), but perhaps this is no longer the case. Particuarily now that there is the LGD option.

[QUOTE=LaurV;381057]To be more specific a third option for "type" would also be ok. Like "I don't mind what you give me to do" hehe..

To avoid any confusion, all my cuda cards are on "Let GPU72 decide" and I understand by that the fact that GPU72 can give me to do whatever it considers necessary, bitlevel, range, [U]and type[/U], but the opencl cards are on 73 bits max, because mfakto 74++ bit kernels are lousy (i.e. much slower compared with lower-bits kernels).[/QUOTE]

OK, then...

With everyone's understanding, in the near future the LLTF "Let GPU72 Decide" option may occationally give DCTF candidates for TF'ing.

Anyone who doesn't want to ever do DCTF'ing should choose LLTF'ing "What Makes Sense" with a pledge to 74 (or lower). WMS will never change the worktype, nor set the pledge level higher (it may lower it, however, if needed (not likely)).

Conversely, those who only want to do DCTF'ing should choose the DCTF Worktype; any option. The DCTF worktype will never be upgraded to LLTF.

Does that make sense to everyone? Any mistakes in my thinking?

I thought the whole thing was fairly obvious. Go where they gots the things you wants.

"Why do you rob banks?"
"Because that's where the money is."
-Willie Sutton

Sounds good to me.

The one option I'd like to see added, not that it affects me personally though, is a way to tell "Let GPU72 Decide" that the card, i.e. an AMD, has a penalty for factoring beyond a certain bit depth. "Let GPU72 Decide" ignores the max bit-depth when this may not be ideal. I don't know what the best solution there would be.

[QUOTE=Mark Rose;381075]The one option I'd like to see added, not that it affects me personally though, is a way to tell "Let GPU72 Decide" that the card, i.e. an AMD, has a penalty for factoring beyond a certain bit depth. "Let GPU72 Decide" ignores the max bit-depth when this may not be ideal. I don't know what the best solution there would be.[/QUOTE]

I have been thinking about this myself for some time...

In the short term, those with AMD/Intel GPUs shouldn't use the "LGD" option, but instead use "WMS" (or any option other than LGD) with a defined bit-level low enough that it's optimal for OpenCL's cross-over points.

Longer term, I can add an option to the forms for different GPU types, but this could explode in complexity -- CUDA vs OpenCL would be trial, but even within the CUDA space the cross-over points differ between different chip classes.

Worth thinking about, and eventually implementing. But I'm about to enter a three month contract with /seriously/ hard deadlines, so it won't happen before December.

Perhaps before then a Wiki / Forum page could be created with advice to users which I could easily link to from the assignment forms.

[QUOTE=chalsall;381233]but even within the CUDA space the cross-over points differ between different chip classes.[/QUOTE]

I've certainly noticed a difference in the chips. I find that GTX 760 (GK104) and GT 430's (GF108) get basically the same TF GHz-d/d whether I'm doing 35M -> 71 or 68M -> 74 and use the same tunings in mfaktc.ini for best results.

My GTX 580 (GF110) is completely different. I get 5% more GHz-day/d doing 35M -> 71 (440) versus 68M -> 74 (420), and the ideal mfaktc.ini tunings are different.

[QUOTE=chalsall;381053]

To put on the table though, I already do this with LaurV and Jerry (with their knowledge and permission) -- the LLTF form will assign DCTF iff needed. If you (and/or anyone else) is willing, I can in the short term implement this individually. Please PM me with authorization to send your clients DCTF when needed (this will only be done when LLTF LGD is set as the option, and not very often).

[/QUOTE]

@Chalsall: I tried sending you a PM authorizing DCTF in this fashion but got an error:

[CODE]chalsall has exceeded their stored private messages quota and cannot accept further messages until they clear some space.[/CODE]

I made this suggestion in another thread a while ago.

The "Completed Assignments" only shows up to the 1,000 most recent assignments. In this case, the text "1000 Assignments completed" is kind of misleading. My suggestion is this: if the user has completed more than 1,000 assignments, then the text should read: "1000 most recently completed assignments."

This one is quick: would it be possible to adjust the links on the worker factors found page to include &full=1?

[QUOTE](Pete: we still miss you!!! :sad:),[/QUOTE]

Sad agreement here.

[url]http://www.gpu72.com/graphs/dctf/week/[/url] currently shows a line for TF to 69, but all that work is done. Perhaps it should be replaced with TF to 73 and above?

[QUOTE=Mark Rose;401770][url]http://www.gpu72.com/graphs/dctf/week/[/url] currently shows a line for TF to 69, but all that work is done. Perhaps it should be replaced with TF to 73 and above?[/QUOTE]

Some day. Currently waiting [URL="https://www.gpu72.com/graphs/dctf/year/"]another four months or so until the "69'ers" fall off the graph[/URL]... :smile:

[QUOTE=Mark Rose;401203]This one is quick: would it be possible to adjust the links on the worker factors found page to include &full=1?[/QUOTE]

Bump.

[QUOTE=Mark Rose;402193]Bump.[/QUOTE]

Thanks for the "ping". Done.

[QUOTE=chalsall;402197]Thanks for the "ping". Done.[/QUOTE]

Thank you :)

Evidently [url]http://www.mersenne.info/[/url] has not been reflecting properly updated statistics for more than a week now.
Is that ever to be made to work again?

[QUOTE=snme2pm1;403606]Evidently [url]http://www.mersenne.info/[/url] has not been reflecting properly updated statistics for more than a week now.
Is that ever to be made to work again?[/QUOTE]

Please be aware that Mersenne.info is quite separate from all other systems.

And, to be honest, I don't know if it will ever be back online.

[QUOTE=chalsall;403609]Please be aware that Mersenne.info is quite separate from all other systems.

And, to be honest, I don't know if it will ever be back online.[/QUOTE]

It would be sad to lose the only available facility to observe from various perspectives the thrust of work.
Maybe MadPoo can be encouraged to introduce some of the techniques that you have well provisioned until now.
I was happy to visit mersenne.info often.
G

[QUOTE=snme2pm1;403616]It would be sad to lose the only available facility to observe from various perspectives the thrust of work.

I was happy to visit mersenne.info often.
G[/QUOTE]

+ 1

[QUOTE=snme2pm1;403616]It would be sad to lose the only available facility to observe from various perspectives the thrust of work. Maybe MadPoo can be encouraged to introduce some of the techniques that you have well provisioned until now.[/QUOTE]

Please believe me, I'm as pissed off (if not more so) than others about this. I use(d) this tool / data source for my strategic planning.

Unfortunately the machine behind the work took more of a hit than expected. Also, I'm rather seriously busy at the moment, so I can't immediately jump in and start rebuilding the delta datasets which are fundemental to its functionality.

Next week I will reach out to Madpoo to discuss how to bring this back online.

[QUOTE=lycorn;403651]+ 1[/QUOTE]
another +1 from me

[QUOTE=chalsall;403660]Please believe me, I'm as pissed off (if not more so) than others about this. I use(d) this tool / data source for my strategic planning.

Unfortunately the machine behind the work took more of a hit than expected. Also, I'm rather seriously busy at the moment, so I can't immediately jump in and start rebuilding the delta datasets which are fundemental to its functionality.

Next week I will reach out to Madpoo to discuss how to bring this back online.[/QUOTE]

The Primenet server itself is a GIMPS resource... if they approved it we could look at hosting some of that mersenne.info functionality, but it would depend somewhat on the underlying setup. Primenet runs PHP over IIS, with a MSSQL back-end. The code that makes up all the mersenne.info "goodness" could, theoretically, just be integrated into new mersenne.org reports, looking directly at the data.

If that setup is already PHP'ified, then it becomes simpler for sure.

New data structures or analysis on the back-end required to make it work could be more involved.

There could be other possible solutions to getting the useful bits of mersenne.info going somewhere, but I imagine it comes down to the complexity and whether it'd be an approved use of the Primenet server. That's above my pay grade. :smile:

[QUOTE=chalsall;403660]Please believe me, I'm as pissed off (if not more so) than others about this. I use(d) this tool / data source for my strategic planning.

Unfortunately the machine behind the work took more of a hit than expected. Also, I'm rather seriously busy at the moment, so I can't immediately jump in and start rebuilding the delta datasets which are fundemental to its functionality.

Next week I will reach out to Madpoo to discuss how to bring this back online.[/QUOTE]

Does it seem too hard to migrate the best aspects to PrimeNet?
I vaguely recall that Aaron intended to facilitate a means to grab any desired piece of history, but perhaps that doesn't do what you want.
In the absence of re-hosting some most useful aspects, would it not be more favourable to have mersenne.info kill the sick database and start afresh?
Is it not better to see close to correct yesterday information, with little history, in contrast to months old information purporting to be yesterdays numbers?
Edit: Current data with the absence of any history prior to yesterday would be a major improvement.

[QUOTE=snme2pm1;407728]Does it seem too hard to migrate the best aspects to PrimeNet?[/QUOTE]

To put on the record, I would very much appreciate it if this was available via PrimeNet.

Mersenne.info was mostly built for myself, and I exposed it for those who might be interested.

Sadly, the machine behind it died.

How much of work is to make a new type of report? The user requests a range (there could be some defined ranges and the user puts an offset) and the report involves how many exponents are left and some table with their bitlevel. This should not be that much of server load (query+some post-processing), considering also that not everyone is interested.

[QUOTE=bloodIce;407865]How much of work is to make a new type of report? The user requests a range (there could be some defined ranges and the user puts an offset) and the report involves how many exponents are left and some table with their bitlevel. This should not be that much of server load (query+some post-processing), considering also that not everyone is interested.[/QUOTE]

More work than I have time for.

Perhaps others have cycles.

[QUOTE]More work than I have time for.[/QUOTE]
I have meant that mersenne.org could have such report after a direct database query, I do not understand why you should do it, chalsall?

[QUOTE=bloodIce;408000]I have meant that mersenne.org could have such report after a direct database query, I do not understand why you should do it, chalsall?[/QUOTE]

That's exactly my point as well.

[QUOTE=chalsall;408040]That's exactly my point as well.[/QUOTE]

For myself it might be a bit before I'd have any chance to work on adding a new report (vacation schedule and some upcoming work projects). I don't know if George or James would be up for it any sooner.

Plus I think I was a little confused about what was being asked for. :smile:

I don't really have a clue what's being asked for, and I'm not that smart when it comes to writing stuff to query the mersenne.org database, we speak different languages. :unsure:

It would be nice to have something similar to the (broken) Gimps Visualization tool at [url]http://www.mersenne.info/trial_factored_tabular_data/0/0/[/url]

There you can see the progress of trial factorization on Mersenne numbers without known factor. You can see for different ranges of the exponents from 0M - 1000M the trial factoring done expressed in bits. You can zoom in by clicking in the numbers on the table.

Yes. On the provided link, you click "change" on the left side on the "views" tag and click "3 months ago" or "6 months ago", i.e. a date which precedes the time when the things became totally broken. And you can see what we need. Moreover, there is no much code to write, as Chris has already the old code, and I believe he has nothing against passing it over to the new guy :razz:, so it is only to see where it was broken, and repair the links/db queues -- the things started to go wrong when the primenet started moving around from here to there, many other things (except the visu site) were broken too, remember, also things on James' site, etc. Methink is the same situation. A new <td> or <div> in the new layouts, and the spider (which was parsing the old primenet pages right) went MIA. Part of it may be solved by patches in the primenet webpage. (dream on!)

Well, it should be very nice to have it back. I would put some effort into it myself, but unfortunately I am not a web programmer, I can understand the things I read, but I have no experience in doing them myself...

It actually just occurred to me that I should already have this kind of parsed data (count by range and factored bits, by date) available on mersenne.ca, since I need that data to generate the [url=http://www.mersenne.ca/graphs/]graphs[/url] that I do.

I'll take a peek at the data and see if I can come up with a poor approximation of mersenne.info's tables.

[QUOTE=James Heinrich;408171]I'll take a peek at the data and see if I can come up with a poor approximation of mersenne.info's tables.[/QUOTE]

And... I would be happy to provide you with the code which generates the tables, graphs etc. It's in Perl though; I know you and I have a disagreement about what the "P" stands for in LAMP (:smile:), but it should be trivial to host directly on a LAM[P|P] stack, or port...

To be perfectly honest, the rending side of things wasn't / isn't the issue -- it was the back-end server here in Barbados which did the spidering which suffered an unfortunate early death.

P is for Python, no?

[QUOTE=Mark Rose;408175]P is for Python, no?[/QUOTE]

[url]https://en.wikipedia.org/wiki/LAMP_(software_bundle[/url]) suggest it's interchangeable nowadays but used to stand for : Linux operating system,Apache HTTP Server,MySQL relational database management system and PHP. but the picture on the right of the page shows perl, python, and PHP, mysql etc.

[QUOTE=Mark Rose;408175]P is for Python, no?[/QUOTE]

Just a small rant, if I may...

While I /can/ code in Python, I seriously dislike indentation being used for code structure. That (at least in my mind) is what curly brackets are for.

"I sentient programmer. You deterministic compiler. Do what I tell you to do, rather than forcing me to use an IDE to indent the code to YOUR liking. I accept the responsibility if you do /EXACTLY/ what I tell you to do.

(Please keep in mind, I'm old school. I still use VIM as my primary editor, and sometimes dream in C. When I want to conditionally remove a section of code by using "if (CONDITION) { [MANY LINES OF CODE] }" I feel I should be able to without the compiler telling me "Tut tut! You didn't indent to my satisfaction. More tea dear while you try again?")

I actually like Python's pickiness with indentation. No more arguments about where the curly braces go, or problems when someone has a three-space tabulator and indents things to match with spaces. I also use VIM. :)

[QUOTE=James Heinrich;408171]It actually just occurred to me that I should already have this kind of parsed data[/QUOTE]And upon closer inspection I realize I have something that's almost-but-not-quite that: I have data on the distribution of bitsize of known factors, not the distribution of TF for unfactored exponents. I could expand my data-collection, but it would make more sense to just run this off mersenne.org and pull from the real data. Perhaps when Aaron has some free time he can help me figure out the best way to pull that kind of data out, I can assemble a front-end once the data is available.