Assignment Problem?
 

I have a machine set to TF to low limits and was given several 450M and 469M exp's to TF from 64 to 65. I'm wondering if there was a glitch and primenet gave out some of these assignments twice cause I keep getting not needed on the results, and hate to be duplicating work.

[code]
[Comm thread Feb 6 04:17] Result was not needed. TF on M450491927, sf: 64, ef: 65
[Comm thread Feb 6 04:17] CPU credit is 0.0083 GHz-days.
[Comm thread Feb 6 04:17] Result was not needed. TF on M450491929, sf: 64, ef: 65
[Comm thread Feb 6 04:17] CPU credit is 0.0083 GHz-days.
[Comm thread Feb 6 04:27] Result was not needed. TF on M450492247, sf: 64, ef: 65
[Comm thread Feb 6 04:27] CPU credit is 0.0083 GHz-days.
[Comm thread Feb 6 04:55] Result was not needed. TF on M450492277, sf: 64, ef: 65
[Comm thread Feb 6 04:55] CPU credit is 0.0083 GHz-days.
[Comm thread Feb 6 05:23] Result was not needed. TF on M450492487, sf: 64, ef: 65
[Comm thread Feb 6 05:23] CPU credit is 0.0083 GHz-days.
[Comm thread Feb 6 05:50] Result was not needed. TF on M450492587, sf: 64, ef: 65
[Comm thread Feb 6 05:50] CPU credit is 0.0083 GHz-days.
[Comm thread Feb 6 06:19] Result was not needed. TF on M450492593, sf: 64, ef: 65
[Comm thread Feb 6 06:19] CPU credit is 0.0083 GHz-days.
[Comm thread Feb 6 06:47] Result was not needed. TF on M450492619, sf: 64, ef: 65
[Comm thread Feb 6 06:47] CPU credit is 0.0083 GHz-days.
[Comm thread Feb 6 06:49] Result was not needed. TF on M469457977, sf: 64, ef: 65
[Comm thread Feb 6 06:49] CPU credit is 0.0080 GHz-days.
[Comm thread Feb 6 06:56] Result was not needed. TF on M469530781, sf: 64, ef: 65
[Comm thread Feb 6 06:56] CPU credit is 0.0080 GHz-days.
[Comm thread Feb 6 07:02] Result was not needed. TF on M469535623, sf: 64, ef: 65
[Comm thread Feb 6 07:02] CPU credit is 0.0080 GHz-days.
[Comm thread Feb 6 07:09] Result was not needed. TF on M469550189, sf: 64, ef: 65
[Comm thread Feb 6 07:09] CPU credit is 0.0080 GHz-days.
[Comm thread Feb 6 07:15] Result was not needed. TF on M450492631, sf: 64, ef: 65
[Comm thread Feb 6 07:15] CPU credit is 0.0083 GHz-days.
[Comm thread Feb 6 07:15] Result was not needed. TF on M469587029, sf: 64, ef: 65
[Comm thread Feb 6 07:15] CPU credit is 0.0080 GHz-days.
[Comm thread Feb 6 07:43] Result was not needed. TF on M450492787, sf: 64, ef: 65
[Comm thread Feb 6 07:43] CPU credit is 0.0083 GHz-days.
[Comm thread Feb 6 08:11] Result was not needed. TF on M450493037, sf: 64, ef: 65
[Comm thread Feb 6 08:11] CPU credit is 0.0083 GHz-days.
[Comm thread Feb 6 08:39] Result was not needed. TF on M450493061, sf: 64, ef: 65
[Comm thread Feb 6 08:39] CPU credit is 0.0083 GHz-days.
[Comm thread Feb 6 09:08] Result was not needed. TF on M450493079, sf: 64, ef: 65
[Comm thread Feb 6 09:08] CPU credit is 0.0083 GHz-days.
[Comm thread Feb 6 09:35] Result was not needed. TF on M450493231, sf: 64, ef: 65
[Comm thread Feb 6 09:35] CPU credit is 0.0083 GHz-days.
[/code]

[QUOTE=bcp19;288451]I have a machine set to TF to low limits and was given several 450M and 469M exp's to TF from 64 to 65. I'm wondering if there was a glitch and primenet gave out some of these assignments twice cause I keep getting not needed on the results, and hate to be duplicating work.

[code]
[Comm thread Feb 6 04:17] Result was not needed. TF on M450491927, sf: 64, ef: 65
...
[/code][/QUOTE]

Could be...or could be a case of someone seeing them undone and grabbing them manually at the same time as you got them assigned automatically.

Depends when you were assigned them...
For example this shows who completed the first one...Feb 01.
[url]http://www.mersenne.org/report_exponent/?exp_lo=450491927&exp_hi=10000&B1=Get+status[/url]
...which also depends on when this user was assigned them.
ie who was first.

[QUOTE=petrw1;288453]Could be...or could be a case of someone seeing them undone and grabbing them manually at the same time as you got them assigned automatically.

...which also depends on when this user was assigned them. ie who was first.[/QUOTE]

Third possibility (which petrw1 alluded to but didn't explicitly state): the user who completed the work was legitimately assigned them, but didn't check in for two months. PrimeNet then reassigned them to you, and then the original assignee submitted the work before you did.

Only George or Scott (or possibly James) could say for sure.

[QUOTE=petrw1;288453]Could be...or could be a case of someone seeing them undone and grabbing them manually at the same time as you got them assigned automatically.

Depends when you were assigned them...
For example this shows who completed the first one...Feb 01.
[URL]http://www.mersenne.org/report_exponent/?exp_lo=450491927&exp_hi=10000&B1=Get+status[/URL]
...which also depends on when this user was assigned them.
ie who was first.[/QUOTE]

I was given the assignments on 1-26. I'd had the days of work to queue a bit high when I first got them got them, and still have several waiting. I just used the Exponent Status and selected to show mine, and the first 100 of them listed so far are completed and list me as having them assigned 1-26.

[QUOTE=bcp19;288460]I just used the Exponent Status and selected to show mine, and the first 100 of them listed so far are completed and list me as having them assigned 1-26.[/QUOTE]

That's because the PrimeNet database is reporting what it knows -- that they've been completed, *and* that they are still assigned to you.

[QUOTE=chalsall;288461]That's because the PrimeNet database is reporting what it knows -- that they've been completed, *and* that they are still assigned to you.[/QUOTE]

I just unassigned all of them as they all seem to have been completed and I was given a bunch of 690M exp's, so no more redoing completed work.

[QUOTE=bcp19;288462]I just unassigned all of them as they all seem to have been completed and I was given a bunch of 690M exp's, so no more redoing completed work.[/QUOTE]

It might be a nice addition to PrimeNet for situations like this that the current assignment's page includes a warning where one or more current assignments "aren't needed".

This wouldn't help for those who "install, configure and forget", but for those who review their assignments regularily it would semi-automate this avoidance of duplication of work.

Edit: Actually, thinking about this a bit more... I wonder why PrimeNet / Prime95 (or mprime) didn't automatically unassign the work (and remove it from the worktodo.txt) file. How often do you have your client(s) check in with PrimeNet (or was this mfakt*)?

[QUOTE=chalsall;288465]It might be a nice addition to PrimeNet for situations like this that the current assignment's page includes a warning where one or more current assignments "aren't needed".

This wouldn't help for those who "install, configure and forget", but for those who review their assignments regularily it would semi-automate this avoidance of duplication of work.

Edit: Actually, thinking about this a bit more... I wonder why PrimeNet / Prime95 (or mprime) didn't automatically unassign the work (and remove it from the worktodo.txt) file. How often do you have your client(s) check in with PrimeNet (or was this mfakt*)?[/QUOTE]

Due to the quantity of assignments it had (I had been running low level P-1's from James' site and maxexponents was set to 1500), it was set to report every 7 days and had last reported on the 31st.

I am running into the same issue, but am using manual assignments. I obtained 12 TF assignments from 119000000 -> 119999999 via the web page while signed on to my ID. Primenet put up 12 lines to copy into my worktodo file. I did so and cycled the program (the only way I could get the program to recognize the new worktodo file). It sent out estimated completion dates on all 12. When it finished the first one, with no factor, I got the "Result was not needed" message. I checked the "Primenet summary," and was puzzled to find that while the number of assigned exponents went from 12 to 11, the number of P-1 available went from 1 to 2. So if I was doing duplicate work, I would not have expected the P-1 to go from 1 to 2. Now, it just finished the next one with no factor. Again, the same behavior on the summary screen. Number of assignments for the range went from 11 to 10, and the number of P-1 available exponents went from 2 to 3. If something is wrong, I want to abort this before any more time is wasted. Any ideas?

What do you mean by "Primenet summary"? You mean the summary of each exponent, or your list of assignments, or your user account summary? What do you mean by "number of P-1 available"? If you mean what's listed on your assignments page, is it possible you registered them as LLs rather than TF? Perhaps provide a list of the expos?


On a somewhat unrelated note, I keep a copy of my P95 on my flashdrive for when I'm using university computers during my labs, etc.; I set all worker threads to run LMH, but consistently and regardless of my efforts, workers 3 and 4 always get 67M 69-->70 assignments, which are certainly not LMH. I tried setting the preference via the Prime95 menus, setting them manually in prime.txt, and even via PrimeNet but nothing is changing the behavior. (When I unreserve them I merely get similar assignments.)

I'm referring to [url]www.mersenne.org[/url] -> left column -> "Primenet Summary," status of every exponent, assigned, type, etc. For the heck of it, I decided to try some TFs in the 119000000 range. The 10 that are 'assigned' (on the summary page as of 2/25/2012) are the ones left on one of my machines. The number of P-1 available is currently 3 on the 119000000 line on the same summary report.

The contents of my worktodo file:
[Worker #1]
Factor=E317EBDAC4778F29378E75D3EE65D8BF,119585911,71,72
Factor=8B21EBA7256DE326F3F00D8FEF250F42,119646911,70,71
Factor=2F16360D7617635D3C1233BAA9677279,119717911,70,71
Factor=E62E701D4B805402A3F62093F48965D7,119787911,70,71
Factor=31F08B235844DE59946AE26EF4B9772D,119888911,70,71
Factor=0F50E0FB61971D4C64E81C8EF877C5AF,119999927,67,68
Factor=039295BE26E1D5BF51F1B617FF4AE2C6,119111051,66,67
Factor=09F3AEF46E3F06208CF53EDE48C820B2,119111071,66,67
Factor=9ABF1FCD444D85898E39FDB7A38DD6C4,119111077,66,67
Factor=70D8A875850B980B8952745CDCD204EC,119111099,66,67

The machine is working on 119585911 now. I'm trying to figure out if these exponents are already done/all effort wasted/I don't know how to control this process properly. If they are already done, why did I get them on manual assignment? If they aren't, why is the server telling me "Result was not needed" on the prior two?

According to [URL="http://www.mersenne.org/report_exponent/?exp_lo=119585911&exp_hi=10000&B1=Get+status"]this[/URL], at least the first effort will be in vain. I didn't look at the others. You should try and get a hold of "Never Odd Or Even" and find out why they are doing your assignments.

I did check the rest of them, and the ones in the 60's are fine, but the ones in the 70's are all already at 72, all done by "Never Odd or Even", as ckdo said. I would guess that he's not reserving them, because it's something of a pain; not good though, as it leads to stuff like this. Out of curiosity, when do you get these assignments from PrimeNet? NOoE did all his work by 19/02/12, so if you got your assignments after that date, then it's PrimeNet's fault.

Here is my current assignment list (as of 2/26/2012):

[SIZE=2]CPU Name[/SIZE] [SIZE=2]Core[/SIZE] [SIZE=2]Exponent[/SIZE] [SIZE=2]Work Type[/SIZE] [SIZE=2]Stage, %[/SIZE] [SIZE=2]Assigned[/SIZE] [SIZE=2]age
days[/SIZE] [SIZE=2]Updated[/SIZE] [SIZE=2]Next Update[/SIZE] [SIZE=2]Estimated
Completion[/SIZE] [SIZE=2]days
to go[/SIZE]
HP_Phenom149040263 LLLL, 25.10%2012-02-15 02:46102012-02-25 08:022012-02-26 08:022012-03-14 00:2318
HP_Phenom354383429 LLLL, 75.40%2012-01-09 06:39472012-02-25 08:022012-02-26 08:022012-03-09 04:4213
HP_Phenom454401401 LLLL, 75.40%2012-01-09 06:39472012-02-25 08:022012-02-26 08:022012-03-09 05:0413
HP_Phenom254576047 LLLL, 75.20%2012-01-09 06:39472012-02-25 08:022012-02-26 08:022012-03-09 08:5313
Acer_N2701119111051 TF 0.00%2012-02-10 18:47152012-02-25 18:502012-02-26 18:502012-03-10 18:1814
Acer_N2701119111071 TF 0.00%2012-02-10 18:47152012-02-25 18:502012-02-26 18:502012-03-10 21:5214
Acer_N2701119111077 TF 0.00%2012-02-10 18:47152012-02-25 18:502012-02-26 18:502012-03-11 01:2515
Acer_N2701119111099 TF 0.00%2012-02-10 18:47152012-02-25 18:502012-02-26 18:502012-03-11 04:5915
Acer_N2701119585911 TFTF72, 13.30%2012-02-10 18:47152012-02-25 18:502012-02-26 18:502012-02-29 21:094
Acer_N2701119646911 TF 0.00%2012-02-10 18:47152012-02-25 18:502012-02-26 18:502012-03-03 05:507
Acer_N2701119717911 TF 0.00%2012-02-10 18:47152012-02-25 18:502012-02-26 18:502012-03-05 14:299
Acer_N2701119787911 TF 0.00%2012-02-10 18:47152012-02-25 18:502012-02-26 18:502012-03-07 23:0611
Acer_N2701119888911 TF 0.00%2012-02-10 18:47152012-02-25 18:502012-02-26 18:502012-03-10 07:4114
Acer_N2701119999927 TF 0.00%2012-02-10 18:47152012-02-25 18:502012-02-26 18:502012-03-10 14:4514

I did the manual assignment on 2/10/2012 18:47. NOoE finished the first one on 2/12/2012, but I have no idea when it was assigned. I'll unreserve those I have in the 70s, delete them from my worktodo, and restart the software on my machine. It would be nice to know why Prime95 gave me the ones in the 70s, or didn't raise a warning as the program checked in every day.

Strange stuff going on here. I unreserved the first five (those in the 70s) and the available P-1 'available' counter went from 3 to 8 on the primenet summary display. Next, I did the manual assign process for 10 more TFs in the same range. With the 'assigned to me' box checked on the exponent status screen, it appears that I have the next expected step for each one. In other words, it seems to be OK now.

Next question. When the assignments are cut/pasted into worktodo.txt, I have to cycle the program in order for it to detect the presense of the newly pasted 'factor=...' lines. The manual communication -> send new estimated completion dates function doesn't seem to trigger a reparsing of this list. Is there a way to do it within the software?

Sort of. This is one of my gripes about Prime95. There is a feature called worktodo.add, a file which Prime95 checks for every now and again (every few hours or so). When it reads assignments in there, it will pause all the workers, add the new assignments to worktodo.txt, and restart the workers. I have my own reasons for not liking that feature, but for your case it should work fine if you're willing to wait a few hours.

@scubabob
I may not read carefully enough, or my English not good enough, but let me ask again: are you doing all this work on a CPU?

If so, you should be aware that your clocks are anyhow wasted, and you better select some other type of work, as P-1 or Double Checks, or even first-time-LL if your CPU's are not very old. The reason is that all the work you do TF-ing one day with a good CPU, it can get done by an average GPU in two hours, or by a very good GPU in minutes. Novadays TF-ing on CPU is waste of resources.

Of course, everyone is free to select the work type he feels most convenient, but you should have all the information first, and not select just blind, so forgive me if you already know that.

Of course there is another question here, how could you convince the manual assigner to give you only palindromes? I think you got them one by one (otherwise you get a range) without caring of (or knowing) the fact that someone else is working on these palindromes already. Well, most probably you fond out in the hard way :D

[QUOTE=LaurV;291015]The reason is that all the work you do TF-ing one day it gets done by an average GPU in two hours, or by a very good [I][U][B]G[/B][/U][/I]PU in minutes. Novadays TF-ing on CPU is waste of resources.[/QUOTE]

That's an important typo, we don't want to confuse anyone.

[QUOTE=Dubslow;291018]That's an important typo, we don't want to confuse anyone.[/QUOTE]
Corrected! Thanks a lot!

LaurV,

It is a CPU, but let me explain. The machine is a first generation Atom (N270) Acer laptop. Having been cast off long ago, it is now used as a photo/file/web/FTP/IRC server, etc. Since it's always on, and almost always near 0%, I put Prime95 on it. I tried an LL on it and it took 10.5 months at 100% to do one (48,xxx,xxx). It did the P-1 beforehand, but I don't remember how long that went. I had no idea the software could exploit GPUs. I suppose the FPUs in the GPUs are stronger than what Intel and AMD can come up with?

How [I]I[/I] convinced the manual assigner to give me palindromes? :) Are you kidding? I have no clue. Not only that, the numbers don't read the same backwards and forwards. I am not a math/numbers expert, however, so it's probably over my head. To me, these are palindromes: [url]http://www.youtube.com/watch?v=Nej4xJe4Tdg[/url] On my first shot at the manual assigner, I asked for and got 12 TF assignments. On the second most recent shot, I asked for and got 10 TF assignments.

Well, 119585911, 119646911, 119717911, 119787911, 119888911 are all palindromes, aren't they? What is the chance you get 5 like that in 10 random assignments? Zero. I was just curious. It could be that someone worked the palindromes in the area in the past to some higher bits (I did not check, just WAG) and that number of bits matched your requirement (and all the other expos in the area are factored to much lower number of bits). Anyhow, this is not important, and I am not accusing you of anything. Do not get me wrong, any help is welcomed, and if you like to work not only palindromes, but even "crocodilian" numbers too, whatever they would be, then be my guest. Maybe you find some interesting things that other ppl did not.

edit ps: for "software exploiting GPU" we have few threads here on the forum, I would highly advise to read them, and if you have any alike hardware, then to join gpu factoring too. That is a lot of fun!:max:

Wow, I really can't say why I didn't see those earlier. :blush:

I'm been here for years for the fun, like you said. I understand the goal, but just a small percentage of all of the details in getting there. There is no angst/frustration/anger from me towards you or anyone else. I work as a programmer (IBM z/machine -> assembly language), so I have a common interest in keeping CPUs busy, but not in a number theory kind of way.

As far as the useful-but-not-terribly-long thing goes, Double Check LLs are around 28M at the moment, which means that while a 48M expo is something like 90 GHzDays to test, the 28M double check would take ~25 GHzDays to do, which would only be around 2 months on the laptop (and we could always use more DC throughput). Just food for thought.

scubabob,

None of your clocks are wasted, as long as you're cooperating with PrimeNet about assignments. Some folks confuse the worktype on which a particular CPU has maximum efficiency with the only worktype worth doing by that CPU. But like LaurV, I'd encourage you to continue doing what you like best.

P.S. I was an assembly programmer, too, for many years and also had a fondness for maximizing what those wonderful CPUs could did.