Re: Mersenne: P-1 Puzzle
- Original Message - From: Brian J. Beesley [EMAIL PROTECTED] To: Daran [EMAIL PROTECTED]; [EMAIL PROTECTED]; Anurag Garg [EMAIL PROTECTED] Sent: Tuesday, June 11, 2002 8:23 PM Subject: Re: Mersenne: P-1 Puzzle On Tuesday 11 June 2002 06:13, Daran wrote: [... snip ... interesting but non-contentious] Very noticable is the proportion of exponents - in all three ranges - which are not getting a stage two effort at all. 26 out the 85 exponents between 795 and 796000, 24 out of 54 between 1550 and 15505000, 35 out of 57 between 33219000 and 33223000. I do not believe that large numbers of P4 systems are being shipped with just 8MB of RAM! This is true. However the philosophy of the project, correctly in my view, is that the software should not cause noticeable deterioration in the performance of a system when it is being run in the background to normal work. I agree. My remarks were intended to make the case for spinning P-1 off into a separate work type, (yeah, I know, it's difficult to change the server code), and to encourage other readers of this list to consider focussing on P-1 work. [...] The default has to be safe;... Again, I agree. While there will be some people who have made a deliberate decision not to allocate extra memory, in many cases people will simply have accepted the default, which means that some machines which could allocate more memory to stage 2 without adversely affecting the user won't be configured to do this. However that same tendency to accept defaults puts GIMPS programmers under an obligation to set those defaults conservatively. ...IMO the current default memory allowance of 8MB is entirely reasonable, even though it causes P-1 to run stage 1 only for any realistic assignment, and even though _new_ systems are usually delivered with at least 256 MB RAM. Against that is the observation that the basic memory footprint has barely changed in the over three years I've been with the project, while typical system memory has increased by a factor of four or more. A default set to 10% of available memory would allow a 256MB system to perform a modest stage 2 on low assignments in the test range, and on DC assignments, while still using proportionately less memory than three years ago. The effect of this could be further mitigated if the memory dialog included radio buttons to limit further the memory usage to 'minimum' (default), with other options being 'reasonable' and 'desireable', (as described in the helpfile) as well as 'maximum', and 'Do not run'. Thus the default would be to run a minimal stage 2 provided it could be done in 10% of system memory or less. I would consider this to be reasonable and conservative. Running P-1 on a 10 million digit exponent requires in excess of 64 MB memory to be allocated in order to run stage 2 at all. That's a lot to ask as a default! It is. OTOH if the box has 1GB installed, then it's not so much. Regards Brian Beesley Daran _ Unsubscribe list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: P-1 Puzzle
- Original Message - From: Daran [EMAIL PROTECTED] To: [EMAIL PROTECTED] Sent: Monday, June 10, 2002 6:54 AM Subject: Re: Mersenne: P-1 Puzzle This appears to have happened to me at least once. I'll spend some time later today cross-referencing my WTD file against pminus1.txt to see if there are any more I don't need to do. Yesterday morning, shortly after 6 UTC, I reserved 42 DC exponents between 7.08M and 7.74M, i.e. all were recently expired exponents. Of these 27 had the P-1 bit set, all of which were P-1 complete, according to the pminus1.txt file dated June 9. Of the 15 with P-1 bit clear, one was in fact P-1 complete. I also reserved 32 test exponents between 12.7 and 13.8M. Of these, 27 had the P-1 bit set, 5 clear. None of these disagreed with pminus1.txt. Finally I checked 28 DC exponents which I had previously reserved, between 7.23M and 7.74M. All had the P-1 bit clear. (I routinely unreserve any exponents I get with P-1 bit set.) Of these, two were P-1 complete. From this, admittedly small and non-random sample, it appears that about one-third of *reassignments* of expired DC exponents have the P-1 bit clear, and that about one in fifteen of these are incorrect, leading to a redundant P-1 rerun by the new owner. I don't know what proportion of DC assignments are reassignments, but given that P-1 testing takes about 2% of the time to run a DC test, I would suggest that the impact of this problem upon the project as a whole is negligible. However, for anyone processing a lot of exponents, particularly those specialising in collecting reassignments, it might be worth putting together a script to identify these anomalies. With 512MB I probably have considerably more available memory than the average machine doing DCs now. That will be less true for machines doing DCs in the future, and probably not true at all for machines doing first time tests. Inspecting pminus1.txt confirms this. I would do a 796 DC exponent with B1=45000,B2=675000. Ignoring those which were obviously P-1ed as first-time test exponents, or where the tester has chosen to go well beyond the optimal limits, there are very few in this range which have been factored deeper than this, and many which have had stage 1 only or a very limited stage 2. I would do a 1550 test exponent with B1=19,B2=4227500, which is better than average, but not to the same extent as with the DC range. And I can P-1 many DC assignments in the time I'd take to do a single test assignment. I'd do a 10M digit exponent with B1=375000,B2=8156250, still better than average, but there are quite a few that go much deeper. Very noticable is the proportion of exponents - in all three ranges - which are not getting a stage two effort at all. 26 out the 85 exponents between 795 and 796000, 24 out of 54 between 1550 and 15505000, 35 out of 57 between 33219000 and 33223000. I do not believe that large numbers of P4 systems are being shipped with just 8MB of RAM! Regards Daran _ Unsubscribe list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: P-1 Puzzle
On Sunday 09 June 2002 08:22, Daran wrote: I'm currently concentrating exclusively on P-1 work. The primenet server doesn't support this as a dedicated work type, so my procedure is to reserve some DC exponants, imediately unreserve any which have the P-1 bit already set, P-1 test the rest, then unreserve them without doing any LL testing. One problem I have discovered is that the server doesn't always 'recognise' that a P-1 result has been returned. It can take several days before my individual account report removes the * indicating that factoring work is necessary. In these cases I hold on to the exponant until the result is recognised in order to stop the subsequent 'owner' from doing a redundant P-1 check. In other cases, the P-1 result is recognised imediately. Though I'm not looking for P-1 specifically, I have seen something similar on a large number of occasions. My current assignment report - the DC part of which follows - contains a number of examples. 6493831 D 64 3.3 33.8 93.8 07-Jun-02 07:25 06-Jun-02 06:02 cabbage 0 v18 6530189 D 64 2.3 27.8 64.8 08-Jun-02 06:02 07-Jun-02 06:02 nessus-b 266 v19/v20 6672569 D 64 31.3 13.8 73.8 14-May-02 07:43 09-May-02 06:05 cabbage 0 v18 6881321 D 64 6.3 23.8 63.8 06-Jun-02 06:06 03-Jun-02 06:06 nessus-j 332 v19/v20 6972001 D* 64 0.3 14.7 60.7 09-Jun-02 04:02 caterpillar 654 v19/v20 7009609 D 63 394908824.3 9.8 64.8 07-Jun-02 06:04 16-May-02 06:06 nessus-m 266 v19/v20 7068857 D 63 588757825.3 0.8 60.8 06-Jun-02 06:06 15-May-02 06:05 nessus-j 332 v19/v20 7076669 D* 64 561798830.3 3.8 63.8 07-Jun-02 06:02 10-May-02 06:04 nessus-b 266 v19/v20 7099163 D 63 269335914.3 11.8 65.8 09-Jun-02 06:26 26-May-02 06:43 T4070 366 v19/v20 7908091 D 64 308019117.7 15.4 60.4 08-Jun-02 21:12 22-May-02 19:17 broccoli 400 v19/v20 7937717 D 64 235929510.5 7.6 60.6 09-Jun-02 02:04 30-May-02 00:30 caterpillar 654 v19/v20 7938407 D 64 131072010.3 12.3 60.3 08-Jun-02 20:29 30-May-02 04:16 vision.artb 495 v19/v20 7940447 D 64 9.8 16.8 65.8 09-Jun-02 06:24 30-May-02 17:39 Simon1 1002 v19/v20 7951049 D 64 65536 7.5 10.7 60.7 09-Jun-02 04:31 02-Jun-02 00:40 rhubarb 697 v19/v20 6972001 and 7076669 are starred although the fact bits column seems to indicate that both trial factoring to 2^63 and P-1 have been run. This is _definitely_ true for P-1 on 7076669, the fact is recorded on my system in both results.txt prime.log. So far as 6972001 is concerned, the database (dated 2nd June) indicates P-1 has been run to a reasonable depth but trial factoring has only been done through 2^62. My system definitely won't have done any more trial factoring yet, let alone reported anything, since that system is set up with v22 defaults i.e. defer factoring on new assignments until they reach the head of the queue. 7009609, 7068857 7099163 are not starred although the fact bits column is one short. The nofactor Pminus1 databases (dated 2nd July) give these all trial factored through 2^62 Pminus1 checked to B1=35000, B2=297500 (or higher). The P-1 limits seem sensible for DC assignments, but shouldn't these have been trial factored through 2^63 like most of the other exponents in this range? Currently, I have nine exponants 'warehoused' whose P-1 results have been returned but not recognised, the oldest was done on May 14, which is rather longer than I would expect. There's no question that the server has correctly recieved the result, because it is contained in a recent version of the pminus1.zip file downloaded this morning along with another four exponants 'warehoused' from May 20. Three more, whose results were returned on June 3 have not yet been recorded in this file. There is an entry in the file for the last of the nine, returned on June 5, but the limits are much smaller than the test I did. The most likely explanation is this is a previous owner's P-1 result which wasn't recognised before the exponant was given to me. I wonder what happens if you're working like Daran and someone returns a P-1 result independently (either working outside PrimeNet assignments, or perhaps letting an assigned exponent expire but then reporting results); if PrimeNet gets two P-1 results for the same exponent, which does it keep? This is not trivial; e.g. if you get no factors, B1=10, B2=100 and no factors, B1=20, B2=20 there might still be a factor which would be found if you ran with B1=20, B2=100. Also, if the database says that P-1 stage 1 only has been run (probably due to memory constraints on the system it ran on), at what point is it worthwhile running P-1 for the possible
Re: Mersenne: P-1 Puzzle
At 13:18 09/06/02 +, Brian Beesley wrote: 6972001 and 7076669 are starred although the fact bits column seems to indicate that both trial factoring to 2^63 and P-1 have been run. This is _definitely_ true for P-1 on 7076669, the fact is recorded on my system in both results.txt prime.log. So far as 6972001 is concerned, the database (dated 2nd June) indicates P-1 has been run to a reasonable depth but trial factoring has only been done through 2^62. My system definitely won't have done any more trial factoring yet, let alone reported anything, since that system is set up with v22 defaults i.e. defer factoring on new assignments until they reach the head of the queue. 7009609, 7068857 7099163 are not starred although the fact bits column is one short. The nofactor Pminus1 databases (dated 2nd July) give these all trial factored through 2^62 Pminus1 checked to B1=35000, B2=297500 (or higher). The P-1 limits seem sensible for DC assignments, but shouldn't these have been trial factored through 2^63 like most of the other exponents in this range? Regarding the star, I've noticed that if an exponent doesn't need any factoring when you check it out, the star pretty much always stays there throughout the entire test. If an exponent does need factoring, the star usually disappears after you finish it ( Daran's observations being a counterexample ). However, I have noticed situations where there are discrepancies with the factoring bit depths. Specifically: 7063451 D 63 This one has P-1 factoring done ( which I did ), but is only factored to 2^62 ( instead of 2^63 ). The nofactor.cmp file on the GIMPS status page agrees with this. However, my worktodo.ini says DoubleCheck=7063451,63,1, which possibly has prevented me from doing trial factoring that I would have otherwise done. Also when I checked out these exponents they had P-1 done, but needed trial factoring to 2^63: 6518503 D* 63 -- DoubleCheck=6518503,62,1 6528541 D* 63 -- DoubleCheck=6528541,62,1 I did the trial factoring on one machine with Factor=6518503,62, and put them on another machine with DoubleCheck=6518503,63,1. When I finished the trial factoring, the star went away, but the factored depth did not update. I decided to release 6518503 back to the server, and someone else checked it out, and it still looks like 6518503 D* 63, so I fear they may repeat the factoring I did. Was this caused by finishing the factoring with a Factor= line instead of a DoubleCheck= line? Nick Glover [EMAIL PROTECTED] Computer Science, Clemson University It's good to be open-minded, but not so open that your brains fall out. - Jacob Needleman _ Unsubscribe list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: P-1 Puzzle
- Original Message - From: Brian J. Beesley [EMAIL PROTECTED] To: Daran [EMAIL PROTECTED]; [EMAIL PROTECTED] Sent: Sunday, June 09, 2002 2:18 PM Subject: Re: Mersenne: P-1 Puzzle 6972001 and 7076669 are starred although the fact bits column seems to indicate that both trial factoring to 2^63 and P-1 have been run. This is _definitely_ true for P-1 on 7076669, the fact is recorded on my system in both results.txt prime.log. It also should be recorded in the worktodo.ini file for that exponent. Now if you were to unreserve that exponent, then contact the server sufficiently late in the day so that it is immediately reassigned back to you (rather than giving you a smaller one), then check the WTD file, then you'll find that the P-1 bit will have been unset. (This is how I discovered the problem in the first place.) There is a small risk of losing the assignment, if someone checks it out seconds after you unreserve it. So far as 6972001 is concerned, the database (dated 2nd June) indicates P-1 has been run to a reasonable depth but trial factoring has only been done through 2^62. My system definitely won't have done any more trial factoring yet, let alone reported anything, since that system is set up with v22 defaults i.e. defer factoring on new assignments until they reach the head of the queue. I'll bet its worktodo.ini entry also has the P-1 bit unset, meaning that you will do a redundant P-1 unless you manually fix it. What does it give for the factored bits? 62, or 63? 7009609, 7068857 7099163 are not starred although the fact bits column is one short. The nofactor Pminus1 databases (dated 2nd July) give these all trial factored through 2^62 Pminus1 checked to B1=35000, B2=297500 (or higher). The P-1 limits seem sensible for DC assignments, but shouldn't these have been trial factored through 2^63 like most of the other exponents in this range? Again, what does your WTD file say? [...] I wonder what happens if you're working like Daran and someone returns a P-1 result independently (either working outside PrimeNet assignments, or perhaps letting an assigned exponent expire but then reporting results); Or if someone is working like me who hasn't identified the problem. If they unreserve an exponent whose P-1 hasn't been recognised by the server, then the next owner will do another one, with possibly different limits. This appears to have happened to me at least once. I'll spend some time later today cross-referencing my WTD file against pminus1.txt to see if there are any more I don't need to do. This is not trivial; e.g. if you get no factors, B1=10, B2=100 and no factors, B1=20, B2=20 there might still be a factor which would be found if you ran with B1=20, B2=100. Also, if the database says that P-1 stage 1 only has been run (probably due to memory constraints on the system it ran on), at what point is it worthwhile running P-1 for the possible benefit of finding a factor in stage 2? That question generalises. If the database shows that a shallow P-1 has been run, at what point is it worth running a deeper one, and with what limits? Suppose the a priori optimal bounds for an exponent are, for example B1=4,B2=60, but it turns out that only stage 1 has been run to 4. Assuming that it's worth re-running the P-1 at all, it might be better to drop the B1 limit - to 35000, say, and increase the B2 limit. This would reduce the factor-space overlap with the first test. What's missing in all this is any kind of quantitative analysis. In any case, as long as there are exponents which haven't had a P-1 at all, I prefer to stick to them. [...] Daran, my advice would be to concentrate on exponents above the current DC assignment range which have already been LL tested but not P-1 checked, or on exponents above the current LL assignment range which have been trial factored but not P-1 checked, according to the official database (updated weekly - you will need the pminus1, nofactor hrf3 database files, plus the decomp utility to unscramble nofactor). I have these files, but following your advice would undermine my rational for doing this. With 512MB I probably have considerably more available memory than the average machine doing DCs now. That will be less true for machines doing DCs in the future, and probably not true at all for machines doing first time tests. I can work around the problem while staying within the current DC range. It's just an irritation. Regards Brian Beesley Daran _ Unsubscribe list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers