[QUOTE=chalsall;332813]Yeah, I know. This was a Stupid Programmer Error (SPE) (I am the programmer) when I added a requested feature without appropriate testing which caused a regression.

Unfortunately I'm *really* busy at the moment. Please trust me when I say that no information has been lost -- it's just a bug in my SQL (or maybe the Perl) which produces that report.[/QUOTE]

Something is really funny about #12 - the total on the overall page jumped 17,000 a couple of days ago, and does not agree with the sum of work types completed.

[QUOTE=Chuck;332817]Something is really funny about #12 - the total on the overall page jumped 17,000 a couple of days ago, and does not agree with the sum of work types completed.[/QUOTE]

What part of "I made a mistake, and I'm really busy!" isn't clear? :smile:

As I used to tell my Mother when I broke something as a child: "I can fix it!". :wink:

Edit: And to be clear, the problems which are being seen are only on the reporting. The core back-end has not been touched and is stable. No duplication of work is happening.

We understand your current time limitations, so maybe consider this post as informational only. We are certain it can be fixed later on.

We set up MISFIT last night. It looks like it is working, and results that have occurred after the setup show up in GPU72's results. But, prior to setting up MISFIT, we dumped ~1,500 GHz-days worth of work, waited a few hours and unreserved all of our assigned work. (We did not know if MISFIT would play nice with our current worktodo.txt files so we decided to start from a clean slate.)

What may have happened is the GPU72 spider missed the upload and when we unreserved all of our assigned work we actually unreserved the work we had already turned in.

We used the manual submission page at PrimeNet.

Attached it the work we turned in. Hopefully it has not been reassigned to anyone.

Sorry for the trouble!

:redface:

[QUOTE=Xyzzy;333306]Sorry for the trouble![/QUOTE]

Grrr... :wink:

OK, the good news is this can be fixed without too much trouble. I'll have to write a little Perl script to un-unreserve these assignments (won't be until Saturday or Sunday).

The database shows you unreserved them before "Spidy" scanned the ranges in question. (Note that for the 65M and 332M ranges, Spidy only checks twice an hour.) Also, it doesn't appear anyone was assigned the candidates before their TFed depth level was updated. As in, no duplication of work occurred.

:busted:

[QUOTE]You can see that for many people, the graphs become much less readable because of "spikes".[/QUOTE][COLOR="white"].[/COLOR]

[QUOTE=Xyzzy;333362].[/QUOTE]
You may access your "view assignment" page on gpu72, select all, and do "regenerate" the worktodo file. Manually append that to your misfitworktodo.txt file which is in the misfit folder. This would be the easiest, and you still do your assignments, Misfit may do good job in taking care about the duplicates, according with the discussions around (I never tested personally this feature, but Scot and Jerry can give you more details), so you don't need to select them one by one in the assignment page. You are done with 3 clicks and a copy/paste, then put MISFIT to redistribute the work.

Related to the second question, MISFIT plays very nice with work already existent, you can stop/restart misfit without stopping/restarting mfaktX, and nothing bad would be happening, and you still can do everything you used to do before, like manually editing the worktodo files, but do not touch the first line, and be careful not to do that just before some exponent finishes the bit level (otherwise you may duplicate some work if the first line is cached into your editor), etc. But honestly, all the manual work is futile now, with MISFIT.

By setting the report interval to 4, 6, 8 hours (like more times per day, default is 4) you will also avoid those spikes.

[QUOTE=LaurV;333401]... Manually append that to your misfitworktodo.txt file which is in the misfit folder. [/QUOTE]

Thanks LaurV... Just to clarify some functionality. From the MISFIT "Add Work" button you get a form with a text box and a FETCH button. Clicking fetch causes an instant fetch to occur. You can also paste data directly into the text box. There are two benefits when using this form.
[LIST=1][*]File-Locking is applied/handled for MISFITworkToDo.txt during the save event.[*]The data in the text box is checked that each row starts with [B]"Factor="[/B] and error is shown if validation fails[*]Blank rows are tossed[/LIST]

As for frequency of uploads: Every upload puts load on the GIMPS server so I favor less frequent uploads.


thx

[QUOTE=swl551;333423]As for frequency of uploads: Every upload puts load on the GIMPS server so I favor less frequent uploads.[/QUOTE]

On the other hand, not much load.

And we're currently in a "race condition". We need to get back to Primenet as many candidates as possible TFed to at least 73 (and, in an ideal world, P-1'ed).

I would argue it's better for people to submit their results sooner rather than later.

Once an hour wouldn't cause Primenet any distress, and would probably help let it release lower candidates for LLing earlier.

Forgive us for being a little slow, but how exactly do we set up MISFIT to provide what GIMPS needs the most?

We have attached a picture of the tab that we think is set up properly.

We assumed that there would be a button for "Let GPU72 Decide!" and then everything else would gray out, but the other options are still active. Or are they?

:mike:

[QUOTE]Once an hour wouldn't cause Primenet any distress, and would probably help let it release lower candidates for LLing earlier.[/QUOTE]Unless we are doing it wrong, which is very likely, we have observed that using the automatic upload feature has a minimum four hour interval, and since the scheduled tasks are limited to 60 events, they are limited to three hour intervals. (7*8=56)

[QUOTE=Xyzzy;333485]Unless we are doing it wrong, which is very likely, we have observed that using the automatic upload feature has a minimum four hour interval, and since the scheduled tasks are limited to 60 events, they are limited to three hour intervals. (7*8=56)[/QUOTE]

You are not doing anything wrong. MISFIT is designed to prevent overly-aggresive uploaders from hammering GIMPS.

There's what, 200 of us doing gpu trial factoring? If the primenet servers can't handle 150-200 hits per hour, we're in trouble. :)

Off topic though!
I try to do p-1 work on my cpu, and I TF exponents to 74 if they're above 60M as per James' chart. Thus I am confused why, when I go to get p-1 exponents with the highest TF factor, none come back that are factored above 73. What happens there?

[QUOTE=Aramis Wyler;333494]There's what, 200 of us doing gpu trial factoring? If the primenet servers can't handle 150-200 hits per hour, we're in trouble. :)

[/QUOTE]

I know its crazy: All the times the site was down or "could not process" or timed out or extremely slow.

[QUOTE=Aramis Wyler;333494]There's what, 200 of us doing gpu trial factoring? If the primenet servers can't handle 150-200 hits per hour, we're in trouble. :)[/QUOTE]

Indeed. While Scott has done a great job of providing Windows users a way of auto-fetching and auto-submitting results, he's not deeply aware of Primenet's behavior.

If each worker submits once an hour it will not be a problem.

[QUOTE=Aramis Wyler;333494]Off topic though!
I try to do p-1 work on my cpu, and I TF exponents to 74 if they're above 60M as per James' chart. Thus I am confused why, when I go to get p-1 exponents with the highest TF factor, none come back that are factored above 73. What happens there?[/QUOTE]

The TFing and P-1 work is decoupled.

GPU72 keeps 500 of the lowest candidates TFed to at least 73 bits in the cache for P-1'ing.

If you TF to 74, but there are more than 500 candidates at 73 below, those at 74 will be released back to Priimenet to manage as it deems appropriate.

Ah, I see. The server chooses the lowest candidates for p-1'ing. I can see how that would make sense.

Do you think it makes more sense than queing up the 500 most Tf'd candidates? Maybe I am wrong in thinking that the highest TF'd candidates are the best ones for p-1 work.

[QUOTE=Aramis Wyler;333500]Ah, I see. The server chooses the lowest candidates for p-1'ing. I can see how that would make sense.

Do you think it makes more sense than queing up the 500 most Tf'd candidates? Maybe I am wrong in thinking that the highest TF'd candidates are the best ones for p-1 work.[/QUOTE]

If were were ahead of the wavefront and thus able to be picky about the order in which we do them, then arguably yes you are correct.

However, since P-1 is so desperately behind the wavefront, the lower it is, the more likely it is to complete the LL test without P-1 at all; so we do lowest first.

Does that imply that primenet offers up lowest number first, not highest factored first to p95?

[QUOTE=Dubslow;333503]However, since P-1 is so desperately behind the wavefront, the lower it is, the more likely it is to complete the LL test without P-1 at all; so we do lowest first.[/QUOTE]

Indeed.

There are already 7 candidates in the 59M range, and 38 in the 60M range, which have had a LL completed without a P-1 done.

Also, "Spidy" is most efficient at simply reserving the lowest candidates without a P-1 done, since that's what Primenet offers first.

[QUOTE=Aramis Wyler;333504]Does that imply that primenet offers up lowest number first, not highest factored first to p95?[/QUOTE]

Correct.

Well holy crap, that's some pretty critical knowledge right there. I will revise my work assignments. ty.

[QUOTE=Aramis Wyler;333507]Well holy crap, that's some pretty critical knowledge right there. I will revise my work assignments. ty.[/QUOTE]

Thank you.

While James has shown us we should be going to 74, we only (just; barely) have the firepower to go to 73, to stay ahead of the "LLing wave".

And for every candidate taken to 74, we could take approximately two to 73.

Things should calm down in a couple of weeks or so -- as predicted, a "surge" of reservations from Primenet occurred after the announcement of M48(?). Normal reservations are only valid for 60 days -- many of these will be recycled soon.

[QUOTE=chalsall;333509]James has shown us we should be going to 74[/QUOTE]... only examining TF and LL both done on GPU. Since there is a relative abundance of CPUs LL'ing and a paucity of GPUs TF'ing, the balance may need to shift away from the supposed optimal shown on my graph.

[QUOTE=James Heinrich;333515]... only examining TF and LL both done on GPU. Since there is a relative abundance of CPUs LL'ing and a paucity of GPUs TF'ing, the balance may need to shift away from the supposed optimal shown on my graph.[/QUOTE]

Since you are the authority on this subject, where should the GPUs currently be TFing to (knowing we can currently hold our lead going to 73)?

Edit: For the record, I really enjoy working with really smart and motivated people. :smile:

[QUOTE=chalsall;333518]Since you are the authority on this subject[/quote]:surprised I beg to differ on that point!

[QUOTE=chalsall;333518]where should the GPUs currently be TFing to?[/QUOTE]You have a much better perspective on that than I would. You can monitor the balance of all the various available assignments, how many PrimeNet is releasing for LL with what level of TF and all that. That's the real-life data on which TF depth assessments should be based, not some theoretical crossover point based on a non-representative configuration (GPUs used exclusively for TF and LL).

[QUOTE=chalsall;333505]
There are already 7 candidates in the 59M range, and 38 in the 60M range, which have had a LL completed without a P-1 done.
[/QUOTE]
Once [URL="http://www.mersenneforum.org/showthread.php?t=17835"]cudapm1[/URL] is complete and working, I will switch GPUs from LL DC to P-1. Hopefully that will help.

A co-worker of mine has volunteer to keep two of my systems in his house after I move! So, I'll have four 580s running TF and two 3770Ks running P-1. Should be good for ~1800 GhzDays/Day.

This is where the rubber meets the road! Everything will be automated as I won't be able to access them 1900 miles away. He has knowledge of computers so he can help when required, but [I]hopefully[/I] the systems will trouble-free for a very long time.

Right now I have them up and running in my house to work out any problems. MISFIT is working well for everything so I think we're good. One of my 580s is running hot though, so I need to T/S later.

When GPU P-1 arrives, I can log in and switch a couple over, if required. Probably won't need to though as others will switch and we'll need the TF power still.

[QUOTE=flashjh;333530]A co-worker of mine has volunteer to keep two of my systems in his house after I move! So, I'll have four 580s running TF and two 3770Ks running P-1. Should be good for ~1800 GhzDays/Day.[/QUOTE]

That's great Jerry! We can really use the TFing and P-1'ing. :smile:

[Channeling my inner manager]You have, of course, warned him about how his electricity bill will be affected?[/channeling]

[QUOTE=chalsall;333547]That's great Jerry! We can really use the TFing and P-1'ing. :smile:

[Channeling my inner manager]You have, of course, warned him about how his electricity bill will be affected?[/channeling][/QUOTE]

Yes :smile:. Each system uses ~$30 per month. I offered to pay but he was happy to work another deal. I gave him one of my i7 580 systems in exchange for unlimited computer work. I only expect him to have them for a year or so, so it about works out.

EDIT: On a side note, he said that he may start running TF when not playing games, so when I see him Monday I'll ask if he wants help setting it up. Since he'll be hosting my systems, I could even set up MISFIT to work with his system.

[QUOTE=flashjh;333530]Everything will be automated as I won't be able to access them 1900 miles away. He has knowledge of computers so he can help when required, but [I]hopefully[/I] the systems will trouble-free for a very long time.[/QUOTE]

You might want to look into software to remotely control your PCs, anyway.

[QUOTE=Mini-Geek;333597]You might want to look into software to remotely control your PCs, anyway.[/QUOTE]

Yes, I use LogMeIn free. It works very well and allows mfaktc and CUDALucas to be started remotely.

re: MISFIT [STRIKE]updating[/STRIKE] reporting results every 4 hours (there is long discussion after, related to different subject, which I have yet to read)

Why would one wants to report results every hour??? The 332M not many people are there (I see only myself and another one or two), the 35-45M DC will be gone in a month (DCTF type of work will disappear in the future, as is reaching the tail of LLTF, which is tf-ed one bit higher, therefore soon will be no DCTF anymore, gone, kaput, basta!). The most needed work, LLTF, is reaching the bitlevels where one assignment takes a hour or more on a GOOD card.

So, why one would want to report faster then he can finish assignments? Let MISFIT to report every 4 hours. I mean, 6 times per day is pretty quite often! I thought I am the most pettifogger here...

[QUOTE=LaurV;332477] I just need 3-4 days more to be able to point you to [URL="http://www.gpu72.com/reports/workers/saved/"]this link[/URL] and say "see what happens when you mix the saved time of LLTF with the saved time of LMHTF? you come on the second place!" :razz:[/QUOTE]
Told you!

This to remain for posterity now: :razz:
[ATTACH]9560[/ATTACH]

It won't report if there is nothing in the results text file, so if it is set to report hourly and you only finish work in 3 hours, it only reports in 3 hours.

[QUOTE=LaurV;333683]Why would one wants to report results every hour??? The 332M not many people are there (I see only myself and another one or two), the 35-45M DC will be gone in a month (DCTF type of work will disappear in the future, as is reaching the tail of LLTF, which is tf-ed one bit higher, therefore soon will be no DCTF anymore, gone, kaput, basta!). The most needed work, LLTF, is reaching the bitlevels where one assignment takes a hour or more on a GOOD card.[/QUOTE]

OK, let me please refine my comment to mean that those doing nominal LLTFing should report their results within a reasonable time as they are completed. Those doing DCTF or 100M digit ranges don't have as much pressure.

But for the LLTF / LL wave, we are still *really* tight. And please keep in mind that while TFing gets cheaper the higher (by Candidate, not bit level) the faster it gets. This is *not* true for P-1ing nor LLing.

At the end of the day, submitting results to Primenet when they are available, rather than in batches, is going to put very little additional load on Primenet.

[QUOTE=LaurV;333726]Told you!

This to remain for posterity now: :razz:
[ATTACH]9560[/ATTACH][/QUOTE]

:rofl: Our guest has been taking up a great deal of our time (he's one of those who needs to be entertained -- he can't entertain himself. And he thinks that because he's on vacation, everyone around him is)... Additionally, I have a hard deliverable for Monday.

Enjoy it while it lasts -- it won't for too much longer.... :smile:

[QUOTE=chalsall;332802]The averaging will be re-introduced soon, with the option of viewing the "true" graphs for those who want to ensure their graphs exactly match their tabular reports.[/QUOTE]

OK... While my SWMBO (and her (*really*) stupid cousin; our guest) were out to watch Polo, and I had a big machine doing some work on my deliverable for Monday, I decided to go into my "comfort zone" and work with computers and their exacting languages (much better in my mind than having to deal with humans and to worry about their "feelings"... :wink:).

The Individual Workers' reports now default again to the adjusted graphs. The "exact" graphs are available via a click-through in the note at the bottom of the page.

A couple of examples where this really makes a difference: [URL="https://www.gpu72.com/reports/worker/7e6a2e592a37a719fac4f765eb0f6ca8/"]Xyzzy[/URL] and [URL="https://www.gpu72.com/reports/worker/6e67460a77a11a707a665a6270df1a82/"]Oliver (AKA The Judger)[/URL].

As always, please let me know if anyone sees any regressions or SPEs.

:bow:

This is good — although I report several times a day, I still like the averaging.

[QUOTE=Xyzzy;333794]:bow:[/QUOTE]
@chris, double that, :bow::bow:

[QUOTE=LaurV;333803]@chris, double that, :bow::bow:[/QUOTE]

Well, at least I can [URL="https://www.gpu72.com/reports/worker/d0a0a9610115a8227c686f0d8951998b/"]see[/URL] excactly how much P-1 GHZ/days I am producing! :P

[QUOTE=chalsall;333518]Since you are the authority on this subject, where should the GPUs currently be TFing to (knowing we can currently hold our lead going to 73)?

Edit: For the record, I really enjoy working with really smart and motivated people. :smile:[/QUOTE]

[QUOTE=James Heinrich;333523]:surprised I beg to differ on that point!

You have a much better perspective on that than I would. You can monitor the balance of all the various available assignments, how many PrimeNet is releasing for LL with what level of TF and all that. That's the real-life data on which TF depth assessments should be based, not some theoretical crossover point based on a non-representative configuration (GPUs used exclusively for TF and LL).[/QUOTE]
Two or three years ago, I (inter alia) judged that the GPU firepower at the time could TF to 72 and keep COMFORTABLY ahead of the LLwave, which was then ~54M.
I think you will agree that this was an accurate assessment then.
After all, you named your effort at coordination "GPUto72" in acknowledgement.

How much has the GPU firepower increased since then?
Should keeping ahead to 73 or 74 bits really be such a struggle?

David

[QUOTE=davieddy;333845]
...
How much has the GPU firepower increased since then?
Should keeping ahead to 73 or 74 bits really be such a struggle?

David[/QUOTE]

It certainly will be easier by this one thing: Get a few GPU yourself, eh? :davieddy:

[QUOTE=davieddy;333845]After all, you named your effort at coordination "GPUto72" in acknowledgement.[/QUOTE]

Actually, I named it that because it rhymed....

Looking at [URL="https://www.gpu72.com/reports/overall/graph/year/"]the year graph[/URL], the total throughput on gpu72 has increased to about 177% of what it was last year. Keeping ahead of the curve at bitlevel 74 is ~400% of the work of keeping ahead of it at 72.

Now those are hard numbers, and 'comfortably' is not, so it's hard to say that we should or should not be able to do gpu to 74 because we could comfortably do it to 72.

With our current avg throughput of approx 15k ghzdays per day (not counting p-1 because the time is the same at 72 or 74) and taking an exponent from 69 to 74 being about 56.5 ghz days, then we could probably chuck out about 265 a day if we were all doing that. But we have a massive backlog of work getting all the dc tf done, and I don't think 265 is 'well above' the number of LLs primenet is handing out every day.

Me, I want rid of the dc tf. If we could get 60 days ahead of the wave in the ll tf front, I'd love to spend 30 days working the dc tf. Then of course we'd have to switch back to the ll tf till we were back up to 60. But we don't work as a block and that's ok. I factor to 74. Most people factor to 73. A bunch of people just do DCs or do non-wave related work (LMH).

It's getting done. This time next year it will be a non-issue.

[QUOTE=Aramis Wyler;333871]With our current avg throughput of approx 15k ghzdays per day (not counting p-1 because the time is the same at 72 or 74) and taking an exponent from 69 to 74 being about 56.5 ghz days, then we could probably chuck out about 265 a day if we were all doing that.[/QUOTE]

Thanks for your immediate above Aramis.

So everyone knows, over the past 30 days ~284.27 LL candidates were LLed per day. Over the last week, this rate increased to ~305.57 candidates per day. My reading of this is the M58(?) announcement has resulted in a greater number of GIMPS participants.

Over the same 30 days, GPU72 has averaged ~600 candidates TFed to at least 73 bits. Over the same week, this rate has dropped to ~500.

Three important things to keep in mind:

1. There was an expected surge of assignments following the announcement.

2. Historically, only approximately 20 percent of all LL assignments actually complete.

3. Completion rates and assignment rates are decoupled.

While I expect a great many candidates to start to "expire" and be recycled by Primenet in a couple of weeks or so, until then we still have to keep ahead of the assignment wave-front. And even in the last week there were days where Primenet handed out more candidates for LLing than GPU72 completed to at least 73 and released.

In the worst case scenario, we are only ~14 days ahead -- going to 73.

In, lets say, three weeks or so we can come back and decide if we have the firepower to go to 74 before releasing any candidates. Until then, those who want to go to 74 (like you Aramis -- thank you for that as well) are free to do so.

[QUOTE=Aramis Wyler;333871]Looking at [URL="https://www.gpu72.com/reports/overall/graph/year/"]the year graph[/URL], the total throughput on gpu72 has increased to about 177% of what it was last year. Keeping ahead of the curve at bitlevel 74 is ~400% of the work of keeping ahead of it at 72.

Now those are hard numbers, and 'comfortably' is not, so it's hard to say that we should or should not be able to do gpu to 74 because we could comfortably do it to 72.
[/QUOTE]
By "comfortably" I mean two things:

1) TF say ~4M in front of LL, allowing for fluctuation in the rates of both,
and plenty of time to do P-1.

2) Keeping ahead of the LLwave leaves enough firepower for folk to prat about in other regions of the spectrum, as many persist in doing.

Be careful when quoting "hard numbers" - they can be easily refuted.
Here's mine: the TF bitlevel should increment when the exponent increases by a factor of 2[SUP]1/3[/SUP] = 26%.

David.

[QUOTE=chalsall;333894]My reading of this is the M58(?) announcement has resulted in a greater number of GIMPS participants.[/QUOTE]

[smartass]I must have missed all of the discoveries since M#48, then...[/smartass]

[QUOTE=davieddy;333908]Here's mine: the TF bitlevel should increment when the exponent increases by a factor of 2[SUP]1/3[/SUP] = 26%.[/QUOTE]

David... I must ask...

If you are so authoritative, why didn't [B][I][U]you[/U][/I][/B] create GPU72?

Rather than complain about almost everything we choose (and chose) to do?

[QUOTE=kracker;333859]It certainly will be easier by this one thing: Get a few GPU yourself, eh? :davieddy:[/QUOTE]
If that would make a significant difference, it simply goes to prove my point.

Although it natural to be impressed by big iron,
mocking folk for lack of it should be a no-no.

Get a brain.

[QUOTE=ixfd64;333910][smartass]I must have missed all of the discoveries since M#48, then...[/smartass][/QUOTE]

Damn... I mis-typed M58(?). I meant M48(?) :smile:

Those who live by the sword die by the sword.

But we have a lot of fun, and get a great deal of work done, in the meantime.... :wink:

[QUOTE=davieddy;333908]Be careful when quoting "hard numbers" - they can be easily refuted.
Here's mine: the TF bitlevel should increment when the exponent increases by a factor of 2[SUP]1/3[/SUP] = 26%.
[/QUOTE]

I don't know you, daveiddy, but I fear for your well being. Do you cover the clocks in your house lest you learn the time? Maybe red lights and green lights are just the man trying to hold you down, I don't know.

All I did was read a chart, look in my backlog at the ghz hours taken to move a number from 69 to 74 (which I did many times in the 65m range recently) and do some 3rd grade math on how many times the one divides into the other. I challenge you to easily refute any of them.

You returned with a comment about how you felt about bitlevels. Thrilling. You might even be right, but nothing about how you feel refutes anything I said. I simply read the dial on the clock, and did some basic math so you and others wouldn't have to.

[QUOTE=Aramis Wyler;333963]I don't know you, daveiddy, but I fear for your well being. Do you cover the clocks in your house lest you learn the time? Maybe red lights and green lights are just the man trying to hold you down, I don't know.

All I did was read a chart, look in my backlog at the ghz hours taken to move a number from 69 to 74 (which I did many times in the 65m range recently) and do some 3rd grade math on how many times the one divides into the other. I challenge you to easily refute any of them.

You returned with a comment about how you felt about bitlevels. Thrilling. You might even be right, but nothing about how you feel refutes anything I said. I simply read the dial on the clock, and did some basic math so you and others wouldn't have to.[/QUOTE]
For the most part, you should ignore davieddy. He has harped on for well over a year now about the work done by those of us who choose to use GPUs to try and keep ahead of the wavefront. Somewhere in his brain though is a cross-circuit between us being useful and his not being able to be snide about our minor problems. Simple logic shows we(GPU72) have done wonders in making sure the majority of exponents are TF'd to a point where people can expect that most reasonable factors have been found, but we are faulted by davieddy every time a single exponent gets by us, no matter what the reason.

As my dad often says: It's a thankless job (especially where davieddy is concerned) but someone has to do it.

[QUOTE=LaurV;333683]re: MISFIT [STRIKE]updating[/STRIKE] reporting results every 4 hours (there is long discussion after, related to different subject, which I have yet to read)

Why would one wants to report results every hour??? The 332M not many people are there (I see only myself and another one or two), the 35-45M DC will be gone in a month (DCTF type of work will disappear in the future, as is reaching the tail of LLTF, which is tf-ed one bit higher, therefore soon will be no DCTF anymore, gone, kaput, basta!). The most needed work, LLTF, is reaching the bitlevels where one assignment takes a hour or more on a GOOD card.

So, why one would want to report faster then he can finish assignments? Let MISFIT to report every 4 hours. I mean, 6 times per day is pretty quite often! I thought I am the most pettifogger here...[/QUOTE]
I don't understand what you mean here, how will the 35M-45M DC be gone in a month?

[QUOTE=bcp19;333971]I don't understand what you mean here, how will the 35M-45M DC be gone in a month?[/QUOTE]

That was a figure of speech. In the most optimistic case they will be gone in 4-6 months, and in the most pessimistic, in one year and half. But they will be gone. GPU72 shows about 200 days or so, but this number means nothing, it is too low, because not all participants do DCTF, and it is too high because it does not consider that more and more people join the fray, with newer, more performing, GPUs (see the Titan discussion).

The idea was that DCTF [COLOR=Red][B]will be gone[/B][/COLOR], sooner or later. The exponents which are LL-ed now have already TF done [B]one bit higher[/B] than the DCTF would do. So, they [B]will not need DCTF[/B] (except in the case when some 4-5 times faster hardware/software appears, to justify TF-ing "one bit higher" at least, for "only one LL saved" - which now is not justified).

In the future will be no "DCTF" type of assignment.

[QUOTE=Aramis Wyler;333963]You returned with a comment about how you felt about bitlevels. Thrilling. You might even be right, but nothing about how you feel refutes anything I said. I simply read the dial on the clock, and did some basic math so you and others wouldn't have to.[/QUOTE]
My statement about upping the bitlevel when the exponent increases by 2[SUP]1/3[/SUP]assumes your implication that 73-74 takes twice as long as 72-73 for a given exponent.

I didn't say or imply you were wrong.

There are indeed some simple generalizations which can be made.
By using the term "comfortably ahead" I was attempting to avoid cheap remarks from Chalsall and his chums.

No such luck:(

David

[QUOTE=LaurV;333978]GPU72 shows about [STRIKE]200 days or so[/STRIKE],[/QUOTE]145 days from 35M to 46M, including ALL PrimeNet candidates, which GPU72 does NOT yet own.

[QUOTE=bcp19;333970]For the most part, you should ignore davieddy. He has harped on for well over a year now about the work done by those of us who choose to use GPUs to try and keep ahead of the wavefront. +Somewhere in his brain though is a cross-circuit between us being useful and his not being able to be snide about our minor problems. Simple logic shows we(GPU72) have done wonders in making sure the majority of exponents are TF'd to a point where people can expect that most reasonable factors have been found, but we are faulted by davieddy every time a single exponent gets by us, no matter what the reason.

As my dad often says: It's a thankless job (especially where davieddy is concerned) but someone has to do it.[/QUOTE]
David Eddy has been monitoring GIMPS progress for ~8 years and has carefully thought about optimizing the strategy involved.

I find bcp19's remarks highly offensive.
See my "polite fight" with Chalsall for an idea of how to conduct a constructive discussion.

D

[QUOTE=davieddy;333983]David Eddy has been monitoring GIMPS progress for ~8 years and has carefully thought about optimizing the strategy involved.

I find bcp19's remarks highly offensive.
See my "polite fight" with Chalsall for an idea of how to conduct a constructive discussion.

D[/QUOTE]
Good, I wasn't trying to be polite. I've only been a member for a bit over a year, and in that time, I have risen to #6 on the top producers list and to #10 overall since the project started. I have put a lot of time and effort and moeny into this project and I find it highly insulting that you belittle it the way you do. This is a volunteer project, but often times your posts make me feel like you think you are entitled here.

Bluntly, I respect chalsall, I'm starting to respect R.D. since he's calmed down a lot lately, I don't respect you.

[QUOTE=bcp19;334100]Good, I wasn't trying to be polite. I've only been a member for a bit over a year, and in that time, I have risen to #6 on the top producers list and to #10 overall since the project started. I have put a lot of time and effort and moeny into this project and I find it highly insulting that you belittle it the way you do. This is a volunteer project, but often times your posts make me feel like you think you are entitled here.

Bluntly, I respect chalsall, I'm starting to respect R.D. since he's calmed down a lot lately, I don't respect you.[/QUOTE]

I've always had the opinion that even though others insult us, it is no reason to "return the favor".

[QUOTE=kracker;334101]I've always had the opinion that even though others insult us, it is no reason to "return the favor".[/QUOTE]

Very good point. Since when is a community of workers motivated to insult so much. I actually show threads from this forum to people at work as a tool to ensure cross team communication stays positive.

Of course an occasional insult is to be expected. Some, on the other hand, seem to thrive from it.

Good or bad it proves entertaining.

[QUOTE=kracker;334101]I've always had the opinion that even though others insult us, it is no reason to "return the favor".[/QUOTE]
Maybe I am just touchy where davieddy is concerned.

A year or so ago, I was having to search through the database to get an idea of what to work on for my GPU, having to manually create worktodo's since I was working outside the normal areas primenet assigned exponents. When the idea of GPU72 was brought forward and chalsall made it work, it saved me a lot of time and effort in finding the exponents I wanted to work on. Even though I now have tools like MISFIT to aid me, I am still a bit of a throwback, I prefer to manually get my assignments while letting it send my results for me.

During all of this, chalsall has taken a lot of grief, a fair amount of kidding and gotten a lot of praise for his efforts. Sadly, the person who suggested making this system is also the person who has given it the most grief. Herculean efforts were made to not only ensure new exponents handed out were properly TF'd, but to also grab and TF those exponents recycled through primenet as well. Having a basis in programming (I started out going for a CompSci major before joblessness ended my education and I joined the military and ended up an Electronics Technician), I have a very good idea what lengths chalsall has gone to in the creation of GPU72.

We have finally arrived at a point where we are somewhat ahead of the wave, but unfortunately not far enough ahead to guarantee all new LL assignments are "properly" TF'd. To then see "insults lobbed from the cheap seats" while ignoring all the time and effort taken to bring us to the point we are now at is a slap in the face to all of the people who brought us this far.

Maybe the whole problem is that the same thing is repeated over and over like a broken record. GPU72 has helped find over 12,000 factors saving over 4 million GHz days of work (2 years in terms of curtisc effort) yet we are still getting dinged for a single LL "not properly TF'd" being assigned.

In a little over a year, GPU72 has pushed the wavefront almost a year beyond where it would be if there had been no GPU72. Yet all we hear is "you missed one".

Maybe I'm tired of turning the other cheek.

[QUOTE=bcp19;334123]Maybe I am just touchy where davieddy is concerned.[/QUOTE]

Thanks for the kind words above bcp19, but please don't sweat it.

Those who matter know [B][I][U]we're[/U][/I][/B] collectively doing the best we can with the resources available.

Maybe I won't put it into the same words like he did, but I am with bcp here. I don't post in the threads related to subjects I don't know, or to projects I don't contribute to, in spite of the fact that I read all posts (ex: obd, opn, sierpinski, riesel, msieve, crus, boinc, etc). I claim that other people do the same: [B]don't spit out your venom about the work you never do[/B]. At least, doing TF is helping other people find primes, but we do this for fun, and not from altruism. People who only run LL tests like Davieddy (he said it himself so many times) are selfish, are motivated only by EFF's/GIMPS money, or fame, etc, and they should shut up when it come to me, I mean ME, choosing what kind of work I do, and how high I want to TF, if I want. At least, we [B][U]ARE[/U][/B] trial-factoring.

I don't completely think that is right. If I do only LL (and I had no gpu etc) that wouldn't make me selfish, right? As long as I don't complain what others like to do?

[QUOTE=LaurV;334147]
People who only run LL tests like Davieddy (he said it himself so many times) are selfish, are motivated only by EFF's/GIMPS money, or fame, etc,[/QUOTE]

Davieddy does the only sensible work for his CPU.

It is in a sense the least gratifying.
Every 2 months you are informed that Mxxxxxxxx is not prime.

Double checks at least give you the satisfaction of the agreement of the residue with the first test.
However, I consider that the DC front has been pushed too high.
If it were ~ 1/3 of the LL front, your chance per year of finding a prime would be at least comparable to first time tests, and an incentive for new recruits to GIMPS.

[B]Now just shut your bleedin' mouth.[/B]

To say I am sufficiently qualified to discuss the feasible/desirable TF bit level would be an understatement.

David

PS I don't broadcast my CV on "social media".
One reason for this is a vestige of modesty.
And I hope the quality of some of my posts speaks for itself.

[QUOTE=davieddy;334151]To say I am sufficiently qualified to discuss the feasible/desirable TF bit level would be an understatement.[/QUOTE]
I am sufficiently qualified to discuss different subjects, but it would be totally impolite or insane to discuss them in a math-related forum...

Why don't you do some other type of work, which is "more rewarding" and better suited for your old processor? Not necessary gimps, say aliquots. They don't pay do they? I would comment nothing, and really appreciate you if you would do aliquot (or other non-gimps) related work, AND comment about TF and gpu72. I would appreciate your knowledge on the subject.

But advising people to do TF, and what kind of TF to do, and to which bitlevel, but saying in the same time that you only do LL ([U][B]first time[/B][/U] LL on your rubbish computer!) it sounds totally hypocrite[SUP]TM[/SUP], it is like you asking us to provide good, well TF-ed exponents for you, lay the red carpet in front of your feet, so you be able to find your prime, but in the same time willing to sacrifice nothing, to give nothing back to the community.

This is what anger the most people here, not your knowledge. It is your attitude.

[QUOTE=LaurV;334153]I am sufficiently qualified to discuss different subjects, but it would be totally impolite or insane to discuss them in a math-related forum...

Why don't you do some other type of work, which is "more rewarding" and better suited for your old processor? Not necessary gimps, say aliquots. They don't pay do they? I would comment nothing, and really appreciate you if you would do aliquot (or other non-gimps) related work, AND comment about TF and gpu72. I would appreciate your knowledge on the subject.

But advising people to do TF, and what kind of TF to do, and to which bitlevel, but saying in the same time that you only do LL ([U][B]first time[/B][/U] LL on your rubbish computer!) it sounds totally hypocrite[SUP]TM[/SUP], it is like you asking us to provide good, well TF-ed exponents for you, lay the red carpet in front of your feet, so you be able to find your prime, but in the same time willing to sacrifice nothing, to give nothing back to the community.

This is what anger the most people here, not your knowledge. It is your attitude.[/QUOTE]

I think it would help if you read my posts before engaging your gob.
I once said to Chris "You aren't listening".
"No. Not to you" was his reply.

I think we have slightly changed our tune since then.
Perhaps you might consider doing likewise.

D

[URL="https://www.youtube.com/watch?v=ulNSlES1Fds"]Explanation[/URL]

Personally, I find p-1 more rewarding on the cpu than the LL, just because it's needed so badly to keep ahead of the wave. The MaxHighMemWorkers setting in local.txt really lets you do that with onlly a couple gigs of ram. My pc at work only has 2 gigs of ram, has MaxHighMemWorkers at=1, and allocated p95 256MB of memory in the day and 1.5 GB of memory at night. So it crunches stage 1s while I'm at work and stage 2s at night.

Just my 2 cents, regarding cpu work.

[QUOTE=Aramis Wyler;334182]Personally, I find p-1 more rewarding on the cpu than the LL, just because it's needed so badly to keep ahead of the wave. The MaxHighMemWorkers setting in local.txt really lets you do that with onlly a couple gigs of ram. My pc at work only has 2 gigs of ram, has MaxHighMemWorkers at=1, and allocated p95 256MB of memory in the day and 1.5 GB of memory at night. So it crunches stage 1s while I'm at work and stage 2s at night.

Just my 2 cents, regarding cpu work.[/QUOTE]

Note that stage 2 takes longer than stage 1, so this only works well when your "night" is longer than your "day" (for a work PC this should be, but for a home PC maybe not).

[QUOTE=Aramis Wyler;334182]Personally, I find p-1 more rewarding on the cpu than the LL, just because it's needed so badly to keep ahead of the wave. The MaxHighMemWorkers setting in local.txt really lets you do that with onlly a couple gigs of ram. My pc at work only has 2 gigs of ram, has MaxHighMemWorkers at=1, and allocated p95 256MB of memory in the day and 1.5 GB of memory at night. So it crunches stage 1s while I'm at work and stage 2s at night.

Just my 2 cents, regarding cpu work.[/QUOTE]

I must turn off my PC at work :sad:

Luigi

[QUOTE=Mini-Geek;334188]Note that stage 2 takes longer than stage 1, so this only works well when your "night" is longer than your "day" (for a work PC this should be, but for a home PC maybe not).[/QUOTE]


Sure, you're right. I only brought it up because my work machine has low memory - maybe I rambled too much. At home I have it set up differently, with 2 high mem workers most of the time, and 4 when cores go idle every few weeks.

Remedial couse for the culturally deprived
 

[url=http://www.youtube.com/watch?v=a2cnRCCHR1k]Sic 'em Pigs[/url]

When Oliver launched mfaktc, MiniGeek noted that it was ~100 times faster at TF than a CPU.

Most of us realized a few obvious things immediately.

1) It had rendered CPUs redundant for TF
2) GPUs could TF a few bits more

The only question left was "How many more bits?"
This is what I have been addressing.

David

link blocked in the United States by EMI.

[QUOTE=Aramis Wyler;334200]link blocked in the United States by EMI.[/QUOTE]Fascist Pigs.
Canned Heat

[QUOTE=davieddy;334199]The only question left was "How many more bits?" This is what I have been addressing.[/QUOTE]

Not as well, nor as authoritatively, as [URL="http://www.mersenne.ca/cudalucas.php?model=13"]James[/URL]....

[QUOTE=chalsall;334215]Not as well, nor as authoritatively, as [URL="http://www.mersenne.ca/cudalucas.php?model=13"]James[/URL]....[/QUOTE]Did I suggest otherwise?

[QUOTE=davieddy;334216]Did I suggest otherwise?[/QUOTE]
Well, you do seem to disagree with him, right? I mean, your forumula does not mach his, nor even intersect it at any point.

Explicitly, It's hard to say that you're not suggesting that you addressed the question as well or better than James considering you are arguing a case that is not James' and his has been out for weeks.

[QUOTE=chalsall;334132]Thanks for the kind words above bcp19, but please don't sweat it.

Those who matter know [B][I][U]we're[/U][/I][/B] collectively doing the best we can with the resources available.[/QUOTE]
I thought of one more... if GPU72 had not been created to organize the GPU work, we'd still be searching for M48 today.

[QUOTE=bcp19;334228]I thought of one more... if GPU72 had not been created to organize the GPU work, we'd still be searching for M48 today.[/QUOTE]

Someone had calculated how much GPUs have progressed GIMPS in regard to M48, and IIRC it was in the range of <30 days, which means M48 (discovered January 25) probably would have been found by now, anyway. Still, we are making a sizable contribution to GIMPS. :smile:

[QUOTE=bcp19;334228]I thought of one more... if GPU72 had not been created to organize the GPU work, we'd still be searching for M48 today.[/QUOTE]

That's not actually true.

A quick back-of-the-envelope (OK, SQL queries and a scientific calculator) calculation [URL="http://www.mersenneforum.org/showpost.php?p=327756&postcount=1844"]showed that the GPU72 effort helped find M48 ~22 days earlier than it might have otherwise been found[/URL]. I posted this information shortly after the official announcement (when I myself learnt the exponent of the new MP).

We've done good, but we shouldn't claim more than we've actually achieved.

[QUOTE=chalsall;334231]That's not actually true.

A quick back-of-the-envelope (OK, SQL queries and a scientific calculator) calculation showed that the GPU72 effort helped find M48 ~22 days earlier than it might have otherwise been found. I posted this information shortly after the official announcement (when I myself learnt the exponent of the new MP).

We've done good, but we shouldn't claim more than we've actually achieved.[/QUOTE]

Assuming M48 fell into a user who really worked on it instead of leaving it until it expired?

[QUOTE=kracker;334233]Assuming M48 fell into a user who really worked on it instead of leaving it until it expired?[/QUOTE]

Of course. But my spidering records also [URL="http://www.mersenneforum.org/showpost.php?p=327767&postcount=374"]allowed me to speak to this[/URL].

Two Anonymous assignments (60 days each) before finally being completed.

[QUOTE=petrw1;332406]Personal LL total <> top LL total

I think it is related to me unassigning being counted as completed.

Thought I did it right this time. I unassigned from the client only.[/QUOTE]

:bump:
Not nagging ... Just in case it got lost in all the "excitement" lately.

Ha! <snark>Yeah, wtf, Chris! If you've got all this time to blah blah blah on the forums, go fix your code!!</snark>

lol lol lol.

[QUOTE=Aramis Wyler;334243]Ha! <snark>Yeah, wtf, Chris! If you've got all this time to blah blah blah on the forums, go fix your code!!</snark>[/QUOTE]

LOL... Coding needs periods of calmness...

"Talking loudly and quickly" to my suppliers, my employees, my bosses and other forum members in real-time all flows into a different mental state.... :smile:

[QUOTE=chalsall;334245]LOL... Coding needs periods of calmness...

"Talking loudly and quickly" to my suppliers, my employees, my bosses and other forum members in real-time all flows into a different mental state.... :smile:[/QUOTE]
I like this Aramis bloke/ss.

Coding at it's best does indeed involve a certain mind-set.
Not especially well-suited to sociability!

I could clean up this flat in a day or so if it wasn't for Mersenneforum distracting me the whole time.

:smile:

[QUOTE=Bdot;334168][URL="https://www.youtube.com/watch?v=ulNSlES1Fds"]Explanation[/URL][/QUOTE]
:goodposting:
This would be good to watch about once a quarter until it sticks (note to self).

[QUOTE=Uncwilly;334297]:goodposting:
This would be good to watch about once a quarter until it sticks (note to self).[/QUOTE]

Also periodically re-read Gerald Weinberg's "The Psychology of Computer Programming", originally published in 1971. Take special note of "egoless programming".

[QUOTE=Chuck;334300]Also periodically re-read Gerald Weinberg's "The Psychology of Computer Programming", originally published in 1971. Take special note of "egoless programming".[/QUOTE]

And, or... Perhaps [URL="http://en.wikipedia.org/wiki/The_Mythical_Man-Month"]The Mythical Man-Month[/URL]?

It's an old one, but a good one....

GPU to 74!
 

OK, to let everyone know...

Based on my calculations, we should be able to sustainability start going to 74 "bits" at 63M and above. This is based on the current rate of Primenet completions (~312 per day over the last week), our current rate (~500 per day to 73 or above over the last month), and the fact that Primenet will start releasing candidates assigned after the "big announcement" in a couple of weeks or so.

If my calculations are wrong, we have 4,850 candidates already at 73 we can release back when and if needed.

I have adjusted the [URL="https://www.gpu72.com/reports/available/"]Available Assignments[/URL], [URL="https://www.gpu72.com/reports/released_level/"]Released TF Level[/URL] and [URL="https://www.gpu72.com/reports/estimated_completion/"]Estimated Completion[/URL] reports accordingly. Additionally, I have added a new line (>73) to the [URL="https://www.gpu72.com/reports/overall/graph/month/"]Overall System Progress[/URL] graphs, and adjusted the Linear Regression line to take this into account.

I will change the default TF level on the "Get LLTF Assignment" form to be 74 once we've cleared out (or, at least, assigned) all of 62M to 73 (approximately one week).

As always, people are free to take those candidates we hold for P-1'ing below 63M to 74 if they wish, or to not go to 74 when that becomes the default.

[QUOTE=chalsall;334561]and the fact that Primenet will start releasing candidates assigned after the "big announcement" in a couple of weeks or so.
[/QUOTE]

oooohhhhh the drama; the anticipation!!!!

Sounds all good to me...

...exactly what I have been advocating for over 2 years.

[QUOTE=davieddy;334644]...exactly what I have been advocating for over 2 years.[/QUOTE]

Sigh.... :davieddy:

[QUOTE=chalsall;334561]
Based on my calculations, we should be able to sustainably start going to 74 "bits" at 63M and above. This is based on the current rate of Primenet completions (~312 per day over the last week), our current rate (~500 per day to 73 or above over the last month), and the fact that Primenet will start releasing candidates assigned after the "big announcement" in a couple of weeks or so.[/QUOTE]Good to know that the bit level that James judges the optimal is also feasible with current firepower[Quote]As always, people are free to take those candidates we hold for P-1'ing below 63M to 74 if they wish, or to not go to 74 when that becomes the default.[/Quote]But it is not desirable to hoard 3000 assignments at a time. 60 users with 1000 each would do it 3x faster than 20 with 3000.

**********

I note that the expected time to the next prime has dropped from 6 years to 4 in the last year. TFing to 5 more bits accounts for a reduction of 7%.
I expect AVX accounts for the rest.

Each new prime needs ~ 3 times as much computing as its predecessor. For this to occur every 4 years seems a reasonable expectation.

D

Starting to get TF assignments to 74. :smile: