| log in |
|
41)
Message boards :
Number crunching :
Trial Factoring tests
(Message 6133)
Posted 12 Apr 2020 by KEP yes, do this as long as you can, my grandmother died on Friday but no Corona. I'm sorry to hear that, really hope that you got the best of your time with her :) |
|
42)
Message boards :
News :
Trial Factoring - new subproject (beta)
(Message 6121)
Posted 11 Apr 2020 by KEP Also, is there ckeckpoints? If I have a power issue or have to reboot do I loose my current work up to that time or is there a checkpoint saved for it to go back to? Good question, I haven't checked the FAQ. There is checkpointing done every 5 minutes, so in worst case (and then you have to be really unlucky) you will loose 4m59s of work. If I recall correctly, you will only loose any really progress, if you have a crash in stead of a nice shut down of BOINC and the computer in general :) |
|
43)
Message boards :
News :
Trial Factoring - new subproject (beta)
(Message 6094)
Posted 10 Apr 2020 by KEP Maybe that should be in the intro post describing Trial Factoring or in the FAQ. It appears due to the slow move in the amount of unsent tasks, that we have had quite a few users cancel their work, due to them not thinking that they actually progressed. Would such a news or FAQ had made a difference for you STEVE? Would it have made a difference for you Bluestang? |
|
44)
Message boards :
News :
Trial Factoring - new subproject (beta)
(Message 6089)
Posted 10 Apr 2020 by KEP Okay I'll let them run ... Great. Now you made me curious, does the progress bar in BOINC not show the actual progress that mfaktc has made? I'm asking, because I can see that you have actually had progress on all the workunits you aborted. Sorry to break your heart, but the progress looked like this: http://srbase.my-firewall.org/sr5/result.php?resultid=387853299 4.9% complete after 01 h 06 minutes http://srbase.my-firewall.org/sr5/result.php?resultid=387852880 90.3% complete after 1 d 08 h 05 m http://srbase.my-firewall.org/sr5/result.php?resultid=387965108 66.4% complete after 18 h 08 m So did BOINC show 0% complete or what did BOINC show? |
|
45)
Message boards :
News :
Trial Factoring - new subproject (beta)
(Message 6087)
Posted 10 Apr 2020 by KEP I've had some run 18+ Hr's aborted a few. Seems like the ones with GPU72_77-78 name take forever, the GPU72_73-74 named Wu's run shorter Dear STEVE Yes they are very long. We will most likely not see that long workunits for a while. The logic is like follows, if you spent 18 hours on a 77-78 bit at n=105M, then: 77-78 bit = 64800 s 76-77 bit = 32400 s 75-76 bit = 16200 s 74-75 bit = 8100 s 73-74 bit = 4050 s 72-73 bit = 2025 s When you double n to n=210M the test time will look like this: 77-78 bit = 32400 s 76-77 bit = 16200 s 75-76 bit = 8100 s 74-75 bit = 4050 s 73-74 bit = 2025 s 72-73 bit = 1012 s 71-72 bit = 506 s 70-71 bit = 253 s We will most likely startout with the 70-71 bit tests that takes you about 4 minutes per test (these will as we nears nMax take you about 1 minute). The rule of thumb (not 100% exact) is that as bit goes up 1 bit, testing time doubles compared to previous bit. When n doubles testing time is cut in half. We will double 3 times, so a 77-78 bit test will at nMax take about 8100 seconds on your machine. nMax=1,000,000,000 (more or less) nMin=110,000,000 (give and take) Take care, stay safe and please do not abort any tasks, even if they seem to take forever, because credit is adjusted accordingly and you will eventually complete the test, even if you have very old gear :smile: |
|
46)
Message boards :
Number crunching :
Trial Factoring
(Message 6073)
Posted 9 Apr 2020 by KEP That's weird - are you using an edited mfakto.ini or removed TestExponents from it? I have Rebirther might use an edited mfakto.ini - at least if he is using the one I sent him originally. The vectorsize was in that set to 2 (in stead of default), due to very old GPU, but I cannot recall if I made other changes. Maybe Rebirther should change the vectorsize to something new or maybe someone should upgrade mfakto, such that the vectorsize is before testing begins set to optimal vectorsize. Is such an automatical determination and setting of vectorsize part of your optimizations? I'm pretty sure that I did not remove TestExponents from the ini I sent to Rebirther - but again I'm not sure if that is the one he actually uses. |
|
47)
Message boards :
Number crunching :
Trial Factoring
(Message 6063)
Posted 9 Apr 2020 by KEP But I don't understand why Mr. Halsall doesn't want 76-77bit factoring. He did, but there has been some serious miscommunications and now SRBase has been told to remove all references to his personal project and continue on our own. We fortunantly for GIMPs has decided to continue on our own and we are currently in the process of finalising the details on how to reserve work for SRBase. So basically, there may be a temporary shortage of work available, while Rebirther renames the subproject and changes the badges, but after those changes has been made, there should as long as SRBase excist, never be a time without work. Now let's go and chase some factors :) |
|
48)
Message boards :
News :
Trial Factoring - new subproject (beta)
(Message 6056)
Posted 8 Apr 2020 by KEP Then you must remove this from all websites, statistic pages, this is free expressions. You are correct. As long as the name is removed, from the projectlist and from the server aswell the badges that must meet his demands. We are afterall, as you mention allowed the freedom of speech. Keep up the good work and stay strong :) and now back to helping GIMPs :) |
|
49)
Message boards :
News :
Trial Factoring - new subproject (beta)
(Message 6054)
Posted 8 Apr 2020 by KEP There will be a name change of the project due a not permitted name from the website operator. I know it is a bit of a work, but you may also remove the unpermitted name from all message posts where it occurs. Sorry. |
|
50)
Message boards :
Number crunching :
Trial Factoring
(Message 6051)
Posted 8 Apr 2020 by KEP Hello, I'm sorry to inform you, that the workunits you are testing takes about 14,5 hours to complete on your GPU and it appears, by looking at the result file, that the computation was actually completed to 100% before being aborted. Did the BOINC manager not show the workunit as complete before you canceled the tests? |
|
51)
Message boards :
Number crunching :
Trial Factoring
(Message 6017)
Posted 4 Apr 2020 by KEP Came here to check why the new batch is taking me up to 8 hours (1650) to complete and whoa I didn't expect a mix between 10 minutes tasks and 8+ hours task. Should I expect *some* long batches in the future? Yes, it is with Trial Factoring (TF), as is what GPU72 is doing, in order to help GIMPs (Great Internet Mersenne Prime search) boost their search, the same as primetesting, test takes longer the higher the bit. We started out with very small tests, because of testing of setup and software. Now that we have gone in to production phase, it was requested by GPU72 that we took 10000 task that could finish the last needed bit to be trialfactored, so they could be released to a different kind of factoring (P-1) and eventually thereafter to Primality testing. It is more or less (dependant on hardware) such that everytime you increse 1 bit, the task doubles in runtime if the n stays the same. Once everything works, it might be a good idea (if possible) to set up a project, for small tasks and for long running tasks. |
|
52)
Message boards :
Number crunching :
Trial Factoring
(Message 5927)
Posted 29 Mar 2020 by KEP I am cool. You're reading it wrong. I'm trying to help get this figured out. It was said that the app will use both GPUs to run 1 WU and it wasn't and I'm trying to help figure out why. It's nothing against Rebirther, so don't make it so. Okay, then I appologize. Since you found a solution, would you mind posting it here or have you already given it to Reb? ... nice job, sure looking forward to see how much tail you can kick :) |
|
53)
Message boards :
Number crunching :
Trial Factoring
(Message 5920)
Posted 28 Mar 2020 by KEP Dear bluestang I have to ask you to cool down a bit. What you are requesting is not what mfakt(o)(c) was designed for. Having a multiGPU system, doesn't automatically means that the app you use, is able to utilize both GPUs on 1 task. It may be possible in the future, also there may be a tweak that can be done, but as far as I know, mfakt(o)(c) is not multiGPU useable. So in stead of the harsh tone (as I at least read your statements), then please show some support and constructive way to getting this issue solved. This is also brand new territory for Reb, so do not expect that everything will be easy solvable or easily adapted to your needs. Take care and please show some patience and think about how your posts can be read :) Stay safe :) |
|
54)
Message boards :
Number crunching :
Overall project progress?
(Message 4079)
Posted 1 Jan 2018 by KEP Sorry guys, but there is not gonna come a version 0.1 or any complete version of the overall project progress. More information can be found at the CRUS forum on mersenneforum.org |
|
55)
Message boards :
Number crunching :
Overall project progress?
(Message 4076)
Posted 31 Dec 2017 by KEP Version 0.1 of the new statistics will not come on January 1st, but soon after and it will be created using the data available on January 1st. We will not see the huge decrease in overall k's remaining, simply because counting does take long time. We do however for now have a reduction on the k's of about 25,000,000,000,000 k's (this number will increase a lot, as I progresses on counting the various ranges) that are either counted as Trivial Factor k, GFN or MOB. Happy New Year everyone |
|
56)
Message boards :
Number crunching :
Overall project progress?
(Message 3874)
Posted 30 Sep 2017 by KEP Prime conjectures - understood like you are searching the minimum prime-k conjecture like we know it from PSP - correct? Ps. Well it can be mostly automatized, but still hundred thousand of srbsieve ini files need to be created and ran using a batch - however things are now pacing forward and for most conjectures less than 6G (6G of k's seem to be what is optimal to run for each segment of the testings) at least 5% has been tested and fully counted. The good thing about the Trillion plus k's conjectures is that their trivial factors appears to be in a repeat pattern - so in fact all trivial factors and a great number of the MOB's and GFN's will have been excluded in version 0.1 - this means that we will have to maybe test only 50% of all k's that appears untested now. In other words we will see a huge reduction in the k's that need testing and who knows, we might actually see us self reach for the 1% milestone a lot sooner than expected :) |
|
57)
Message boards :
Number crunching :
Overall project progress?
(Message 3862)
Posted 29 Sep 2017 by KEP Yes, it is a little time consuming but mostly because srbsieve has to write the files that can be accurately determined without running all that many ranges - like it is for the trivial factors. But yes, a total sum of GFN and MOB will come, but it will be a few updates into the future - I do however expect to have all the Trivial factors counted under assumption that the Trivial pattern actually persists, when we see version 0.1 of these stats. Most of the work that involves srbsieve, can be automatized, but I still need to create ~100000 srbsieve.ini files. Thanks to predicting the correct amount of trivial factors for R280 and R15, I've been able to already remove approximately 197,000,000,000,000 (197 Trillion) k's that has up until now been part of the progress calculation. Thanks for your feedback :) |
|
58)
Message boards :
Number crunching :
Overall project progress?
(Message 3858)
Posted 27 Sep 2017 by KEP Hi everyone :) I'm gonna have to start creating a new and completely accurate progress calculation. This is starting for sure tomorrow (approximately 24h from now) and it will mean for a long time, that numbers will change, but eventually we will all know exactly how many k's need to be primed and how many is one of following: k with trvial factors k wich is MultipleOfBase (MOB) k wich is GFN All k's with trivial factors, k's wich is MOB and k's wich is GFN (Only Sierpinski side) doesn't need to be tested and therefor should not be part of the overall progress statistics. I'm not sure how this is going to affect the overall progress, but it most likely will make our overall progress increase :) I have thought and desired to create this new stats for a long time and now my head finally seems to have found a way to solve this - using srbsieve. I'm also starting to consider creating a website with the stats, where more floating updates can be presented. I also in the end seemed to need at more modern and easier to work with spreadsheet :) I expect to have version 0.1 ready by January 1st and then version 0.2 a month later and so forward will the increase as we nears/progress to the end of the creation of the new stats by 0.1 I'm open for suggestions on what to have in the stats and maybe a sitename (if I decide to create a webpage) - currently I consider involving following stats: Conjectures proven or tested to n=25K Conjectures proven or tested to n=50K Conjectures proven or tested to n=100K Conjectures proven or tested to n=250K Conjectures proven or tested to n=500K Conjectures proven or tested to n=1M Conjectures proven or tested to n=2.5M Conjectures proven or tested to n=5M Conjectures proven or tested to n=10M Conjectures proven or tested to n=25M Conjectures proven or tested to n=50M Overall progress (both sides) Overall progress Sierpinski side Overall progress Riesel side Overall amount of trivial k's Overall amount of MOB Overall amount of GFN (Sierpinski side) Riesel side amount of trivial k's Riesel side amount of MOB Sierpinski side amount of trivial k's Sierpinski side amount of MOB Sierpinski side amount of GFN Overall amount of primes found Overall amount of Sierpinski primes Overall amount of Riesel primes There may come other additions, but let me hear if you have any suggestions/wishes and consider things to gradually change from now on until the final version where all MOBs and trivials and GFN's are counted on all conjectures (that may be a few years into the future) :) Take care Kenneth |
|
59)
Message boards :
Number crunching :
Overall project progress?
(Message 3538)
Posted 1 Jun 2017 by KEP As of 1st of June 2017 the progress is as follows: 563,674,901,215,566 Total Riesel k's to test (all bases) 142,501,532,757,170 Total Sierpinski k's to test (all bases) 706,176,433,972,736 Total k's to test (both sides, all bases) 563,642,894,123,914 Untested Riesel k's (all not fully tested bases) 142,493,920,181,967 Untested Sierpinski k's (all not fully tested bases) 706,136,814,305,881 Total untested k's (both sides, all not fully tested bases) 645,681 Remaining Riesel k's (all fully and partially tested bases) 200,343 Remaining Sierpinski k's (all fully and partially tested bases) 846,024 Remaining k's (both sides, all fully and partially tested bases) 706,136,815,151,905 Total k's remaining (both sides, untested+remaining k's) 39,618,820,831 Total k's tested or primed 0.0056103289384% of k's tested or primed 99.9943896710616% of k's remaining untested or unprimed Regards KEP |
|
60)
Message boards :
Number crunching :
Overall project progress?
(Message 3534)
Posted 30 May 2017 by KEP Yes there will be an update the first of june, cannot promise when, but there will be an update :) |