Post status updates here

[mataje]

199152: line 4971, size 128, 2^3*3^2, C111

[gd_barnes]

I went up and down the unreserved open sequences comparing them to the DB for 100K-104K. I thought I'd take a hack using only the curves/sieving in the DB for the entire range. Here are some extensions that I was able to do:

100644 i=348 to 361; C98 sz 112
101496 i=1152 to 1167; C94 sz 107 (reserve)
101520 i=843 to 844; C95 sz 104
101862 i=1706 to 1713; C87 sz 102 (reserve)
102480 i=930 to 940; C92 sz 103 (reserve)
102660 i=984 to 997; C101 sz 108
102936 i=1120 to 1122; C100 sz 105
103512 i=1448 to 1450; C98 sz 100


All remaining factors in the entire unreserved 100K-104K range are now "hard" composites.

If I've already reported 1 or 2 of these, please excuse the duplication.

As noted above, I'll go ahead and keep 101496, 101862, and 102480 reserved since their factors aren't too high. None of the rest are reserved.

Karsten, I'll send you a spreadsheet of my reservations/statuses within a few days. I'll still getting a handle on everything here. 10metreh, if you'd like the spreadsheet also, I'll be glad to post it here. It's not easy keeping track of all of this since there is 3 different places where statuses/reservations could be posted.


Gary

[10metreh]

Gary, just out of interest, what are the specs of the system(s) you are running your sequences on? You are doing a lot of work.

Re the spreadsheet: I don't need the statuses (stati?) of the ones that are down in the reservations thread as being reserved by you in the spreadsheet, because I'll update those every weekend anyway. What would be useful is if you could PM me a list of all the work you've done on sequences that are not in the reservations thread.

Oh, and thanks for joining!

[gd_barnes]

LMAO LMAO




I'm not doing anything. I just LOOK like I am. How's the old saying go?: "If you can't dazzle 'em with brilliance, baffle 'em with bull sh__." I certainly won't accomplish the former here so I guess I have unintentionally accomplished the latter. :-)

I'm using 2 cores of a quite slow Athlon 1.6 Ghz dual core laptop and sometimes 1 of them finishes at night and sits idle for a few hours. That's all; I promise.

I think it looks like I'm doing a lot because I'm using the DB to extend things 5-20 indexes but there, once you encounter a hard C86 or higher, you have to stop and run msieve. Also, I've followed up with Clifford to fill in the tougher sequences. In other words, a lot of it is just me point-and-clicking on Syd's excellent DB.

What I'm looking for is the "low hanging fruit" so to speak. That is sequences that are IMHO "a little behind" where they could easily be searched. By that I mean, they're at a point that I can use the DB to extend them easily. Then if I encounter a not-too-tough C86-C90 or so, I put it in my list of stuff to feed into msieve. I've run msieve to factor as high as a C98 but I haven't done many for C95 or greater. In other words, I'm doing the little stuff and leaving the tough stuff. I have a big pile of C87-C97 factors that I'm going to start msieve on here shortly on many of my reservations.

The fill ins that I've done take little time running Aliqueit.exe. It's generally taking 1-2 days to get them filled in because whatever I've reserved generally shows a status with a size of 101 or 102 with a factor of C<95. I hate to say it but using modern machines and programs, it really doesn't take much effort to fill in most sequences up to a size of 100 with an index < ~1500, which is where most 10K-100K sequences are at.

I haven't even successfully run ggfns yet so have avoided the big stuff. I got the cat.exe file from people but I didn't want to spend any more time on that one C99 that I had so many problems with when I was first testing with it.

It's nice to know that it looks like I'm doing a lot of work.


Gary

[10metreh]

Very well done for confusing me into thinking you were using 8 threads of an i7.

[henryzz]

Originally Posted by 10metreh

Very well done for confusing me into thinking you were using 8 threads of an i7.

as far as i know gary doesnt own any i7s yet
he does have >10 Q6600s i think though

[gd_barnes]

Four 2.6 Ghz intel quads and six 2.6 Ghz AMD quads; all Linux. None are being utilized for this. Just my rickety Windows laptop. lol I may put a quad on a couple of interesting sequences at some point. The only down drivers I've caught so far are in fill-in work and they went away before I finished filling them in. I have 2-3 that are barely increasing and 1 that is slightly declining with just 2^2.

[Greebley]

As a side project I have been trying to fill some of the numbers less than 1 million not even in the db (not past 10 digits) up to 80 digits. I wrote a program that runs aliqueit to 80 digits and then moves on to the next so it is mostly automated. Previously, I had to hit ctrl-c and so some stopped at c60 and some went over 100 overnight. Note that these are mostly not going to be in the range worked on yet - i.e. 300k - 1 million, but will eventually have some use.

[gd_barnes]

Great idea Greebley! Is the program easy to use for a newb? If so, could you forward it to me? Is it flexible enough that I could change it to run up to size 105?

What I'd really like to have is something that stops when it hits a "hard" factor of C>=100 as opposed to a specific initial size cutoff. It then goes on to the next sequence.

By "hard" factor, I mean one that is C>=100 where all of the usual number of ECM curves have been run for the specific factor size. That is when I like to stop. Like what you've experienced, sometimes I'll have one run for 4-5 hours after finishing the normal curves on a C102 that I don't want to continue with.


Gary

[gd_barnes]

I'm going to post the status of all of the sequences that I've searched. Several will be duplicates of what I've already posted but I think it will be easier for 10metreh and Karsten to pick them out of here:

11496 i=518, sz 101, C99, -0-
12012 i=1224, sz 115, C102, -0-, finished
12270 i=852, sz 106, C97, +29
12456 i=1835, sz 100, C99, -0-
12960 i=675, sz 106, C101, +18, finished
13350 i=769, sz 107, C98, +24
14022 i=476, sz 101, C98, -0-
14160 i=1049, sz 110, C94, +23
20088 i=466, sz 108, C97, +20
61404 i=1284, sz 101, C99, -0-
100644 i=361, sz 112, C98, +13, finished
101496 i=1167, sz 107, C94, +15
101520 i=844, sz 104, C95, +1, finished
101706 i=2073, sz 105, C95, +40
101862 i=1722, sz 103, C88, +16
102168 i=3049, sz 110, C100, +38, finished
102216 i=800, sz 102, C100, +5, finished
102450 i=739, sz 108, C98, +23
102480 i=951, sz 106, C96, +21
102660 i=997, sz 108, C101, +13, finished
102936 i=1122, sz 105, C100, +2, finished
103248 i=801, sz 136, C109, +3, finished
103512 i=1450, sz 100, C98, +2, finished
103840 i=1144, sz 103, C98, +8
104202 i=1015, sz 105, C100, +21, finished
104400 i=1781, sz 103, C99, +21
105096 i=775, sz 104, C97, +8
105150 i=2117, sz 103, C95, +8
106992 i=745, sz 108, C94, +16
108408 i=591, sz 113, C93, +28
108416 i=1073, sz 105, C97, +45
109134 i=961, sz 105, C100, +3, finished
109710 i=1513, sz 101, C92, +23
111912 i=606, sz 103, C99, +20
112584 i=514, sz 102, C91, +6
113160 i=1613, sz 104, C95, +6
116712 i=1531, sz 104, C98, +12
117348 i=1365, sz 103, C100, +7, finished
117738 i=780, sz 113, C109, +47, finished
122892 i=1763, sz 108, C101, +11, finished
154548 i=696, sz 111, C104, +6, finished
155154 i=869, sz 110, C98, +7
166824 i=2542, sz 101, C89, +31
189390 i=1497, sz 103, C99, -0-
210222 i=295, sz 107, C106, +13, finished
233106 i=1690, sz 108, C102, +28, finished
233320 i=496, sz 108, C101, +19, finished
233324 i=1987, sz 99, C97, +28

Averages of 48 sequences for statistical nerds like me:
103473 i=1150, sz 106, C98, +15

The pluses show how many indexes that I advanced them. Where Clifford had a file, I consider the end of his file as the starting point, which is why several that are < 100K show -0-.

All should be considered reserved unless they say "finished" after them.

Some may have advanced just a few more indexes by the time you read this but most should be close if read within 12 hours.

I thought this would be better than attaching my spreadsheet that has a lot of info. that you don't need.


Gary

[Greebley]

It uses Qt by trolltech which is a op system independent set of utilities that are pretty powerful - a free program you can download. If I get it working well enough it would be a visual program that can run aliqueit and show you the results so far and run more than one sequence, etc. that I could provide for anyone who wanted it - assuming I can get it to work well enough.

Currently, it can take a file or input with lines of the form:
<seq_number> [D<max number of digits>] [T<total time to run>]

So:
276 D80 would run seq 276 to 80 digits.
Multiple lines would run multiple sequences - each could have a (different) D<num> after it as well.
Since there is no way to stop aliqueit at 80 digits the program watches the elf file and kills the process (in essence) when it spots 80 digits to the left of the equals.

It also runs aliqueit with the -s to submit the work when it finishes.

I would like to get the following working:
[c<size of cofactor>] - this may already work - it watches aliqueits log file and looks for a number after the word 'cofactor'. I haven't tried it yet so it may have a bug.

Also not written:
[i<number of iterations>] - so use the -q to run aliqueit say 10 times.
[t<max time for a single iteration>] - use -q and abort aliqueit if a single step takes above a certain amount of time
Also a check for a [up] driver for the sequence and only quit if there is one - otherwise keep going until a driver is hit.
Run using the process priorities aliqueit already has.
Display the elf file in a window so you can look at the sequence so far.
Possible other visual stuff like display the common factor between iterations or the current sequence driver.

Right now it sort-of works, but has some issues:
1) I tried to print out the output from aliqueit into a window, but you get bits and pieces that only give an estimate on what is going on. Mostly it prints the last line. Not sure why its broken.

2) Trying to figure out when to terminate aliqueit is way too expensive in its cpu usage. For example for the digits, it parses the entire elf file over and over again when I really want to read only new information. For this reason it doesn't stop when it should but keeps running several iterations (about a minute) before stopping. Note that parsing the aliqueit.log file in this way would be much worse as that file grows quite a bit.

It runs ok for a dual core windows system though (nothing else tested). I can even run a computer game with a few burps and hiccups.

Since Qt is portable, it should work on unix, but I didn't test.

If you want a copy in its current form, I can give it to you. I would probably test the cofactor stuff and make sure that worked since that would be useful for you. It would probably require deleting your aliqueit.log file on a regular basis (between runs) in its current form.