mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Factoring (https://www.mersenneforum.org/forumdisplay.php?f=19)
-   -   2801^79-1 reservations (CLOSED 27 AUGUST) (https://www.mersenneforum.org/showthread.php?t=13305)

fivemack 2010-04-16 21:24

2801^79-1 reservations (CLOSED 27 AUGUST)
 
After a bit of work, it turns out that the parameters I first thought of are about good enough; so let's get this show on the road.

[code]
n: 77728012992649681652257138568390176857212890117660654903605188181803793527234365966882492843633887478337495016140443699327828796305552459390410906211419468091843928832883028959106125700776135225351872298944331308559262827186539091839208835236997045999057406305939986879
skew: 0.266
c6: 2801
c0: -1
Y1: -1
Y0: 653236457112135142391905382282254342883556401
lpbr: 32
lpba: 32
mfbr: 64
mfba: 64
alambda: 2.6
rlambda: 2.6
alim: 134217727
rlim: 134217727
[/code]

Use the 15e siever, and we'll start by sieving 20 million to [s]320[/s]250 million on the [b]rational[/b] (-r) side.

[b]This will take quite a while[/b]; a range of one million at the top of this region takes about nine days on one thread of an i7/2800 running the 64-bit siever under Linux. It will produce a fair number of relations, but handling those is my concern.

[b][COLOR="Red"]RESERVATIONS CLOSED 27/08/2010[/COLOR][/b]
Reservations:

[code]
26/08 15M-19.8M fivemack DONE 02/09
01/06 19.8M-20M andi47 DONE 04/06
16/04 20M-24M fivemack DONE 25/04
18/04 24M-28M apocalypse DONE 24/04
20/04 28M-30M chris2be8 DONE 02/05
24/04 30M-34M apocalypse DONE 01/05
30/04 34M-38M fivemack DONE 09/05
30/04 38M-40M chris2be8 DONE 12/05
30/04 40M-44M apocalypse DONE 07/05
07/05 44M-48M apocalypse DONE 13/05
07/05 48M-50M chris2be8 DONE 18/05
30/04 50M-52M gimarel DONE 10/05
05/05 52M-56M gimarel DONE 18/05
13/05 56M-60M apocalypse DONE 20/05
14/05 60M-61M richs DONE 21/06
15/05 61M-63M chris2be8 DONE 24/05
18/05 63M-67M gimarel DONE 28/05
12/05 67M-70M batalov DONE 19/05
20/05 70M-74M apocalypse DONE 26/05
22/05 74M-76M chris2be8 DONE 31/05
26/05 76M-80M axn DONE 06/06
26/05 80M-84M apocalypse DONE 02/06
26/05 84M-90M fivemack DONE 03/06
29/04 90M-91M jyb DONE 01/05
18/04 91M-92M jyb DONE 19/04
XXXXX 92M-93M jyb DONE 18/05
29/05 93M-94M axn DONE 08/06
28/05 94M-98M gimarel DONE 10/06
29/05 98M-99M axn DONE 10/06
30/05 99M-100M axn DONE 10/06
30/05 100M-102M chris2be8 DONE 08/06
02/06 102M-106M apocalypse DONE 09/06
06/06 106M-108M chris2be8 DONE 13/06
09/06 108M-112M apocalypse DONE 16/06
11/06 112M-113M andi47 DONE 01/07
13/06 113M-115M chris2be8 DONE 23/06
16/06 115M-119M apocalypse DONE 23/06
17/06 119M-125M fivemack DONE 25/06
07/06 125M-134M fivemack DONE 16/06
12/05 134M-137M batalov DONE 25/05
20/06 137M-139M chris2be8 DONE 01/07
22/06 139M-140M richs DONE 09/08
23/06 140M-144M apocalypse DONE 01/07
24/06 144M-148M gimarel DONE 18/07
29/06 148M-150M chris2be8 DONE 09/07
01/07 150M-154M apocalypse DONE 08/07
05/07 154M-170M fivemack DONE 19/07
08/07 170M-174M apocalypse DONE 19/07
16/07 174M-178M apocalypse DONE 26/07
17/07 178M-180M chris2be8 DONE 25/07
08/07 180M-182M chris2be8 DONE 19/07
20/07 182M-198M fivemack DONE 10/08
23/07 198M-202M chris2be8 DONE 01/08
26/07 202M-206M apocalypse DONE 03/08
30/07 206M-208M andi47 MOSTLY DONE 01/09
01/09 206.25M-207M fivemack
31/07 208M-212M chris2be8 DONE 10/08
03/08 212M-216M apocalypse DONE 10/08
07/08 216M-220M chris2be8 DONE 19/08
10/08 220M-224M apocalypse DONE 17/08
11/08 224M-230M fivemack DONE 16/08
16/08 230M-240M fivemack DONE 27/08
16/08 240M-244M chris2be8 DONE 28/08
17/08 244M-248M apocalypse DONE 24/08
24/08 248M-252M apocalypse DONE 31/08
[/code]
Results

[code]
15- 19 8060344 31/08 fivemack
19-19.8 1650404 02/09 fivemack
19.8-20 836611 04/06 andi47
20- 24 8245584 25/04 fivemack
24- 28 8366872 24/04 apocalypse
28- 30 4180825 02/05 chris2be8
30- 34 8346178 01/05 apocalypse
34- 38 8000250 09/05 fivemack (did some sieving on wrong side)
38- 40 4136909 12/05 chris2be8
40- 44 8210736 07/05 apocalypse
44- 48 8156178 13/05 apocalypse
48- 50 4057673 18/05 chris2be8
50- 52 4041620 10/05 gimarel
52- 56 8026265 18/05 gimarel
56- 60 7977247 20/05 apocalypse
60- 61 1984626 21/06 richs
61- 63 3941410 24/05 chris2be8
63- 67 7870638 28/05 gimarel
67- 69 3897634 18/05 batalov
69- 70 1943211 19/05 batalov
70- 74 7769669 26/05 apocalypse
74- 76 3855657 31/05 chris2be8
76- 78 3832502 06/06 axn
78- 80 3819040 06/06 axn
80- 84 7613647 02/06 apocalypse
84- 90 11286163 03/06 fivemack
90- 91 1870037 01/05 jyb
91- 92 1866216 19/04 jyb
92- 93 1858870 18/05 jyb
93- 94 1860380 08/06 axn
94- 98 7390499 10/06 gimarel
98-100 3683893 10/06 axn
100-102 3673946 08/06 chris2be8
102-106 7283106 09/06 apocalypse
106-108 3635042 13/06 chris2be8
108-112 7209756 16/06 apocalypse
112-113 1804817 01/07 andi47
113-115 3588854 23/06 chris2be8
115-119 7130008 23/06 apocalypse
119-125 10631072 25/06 fivemack
125-128 5286080 13/06 fivemack
128-134 10487475 16/06 fivemack
134-135 1738288 21/05 batalov
135-136 1727485 22/05 batalov
136-137 1730639 25/05 batalov
137-139 3436962 01/07 chris2be8
139-140 1711531 09/08 richs
140-144 6782567 01/07 apocalypse
144-148 6709124 18/07 gimarel
148-150 3337017 09/07 chris2be8
150-154 6594712 08/07 apocalypse
154-162 13040660 19/07 fivemack
162-170 12799828 18/07 fivemack
170-174 6282904 19/07 apocalypse
174-178 6232926 26/07 apocalypse
178-180 3098280 30/07 chris2be8
180-182 3079804 19/07 chris2be8
182-190 12200173 08/08 fivemack
190-198 11992020 10/08 fivemack
198-202 5929744 01/08 chris2be8
202-206 5866800 04/08 apocalypse
208-212 5807956 10/08 chris2be8
212-216 5747945 10/08 apocalypse
216-220 5721865 19/08 chris2be8
220-224 5668818 17/08 apocalypse
224-230 8449423 16/08 fivemack
230-240 13884344 27/08 fivemack
240-244 5488342 28/08 chris2be8
244-248 5444089 24/08 apocalypse
248-252 5408212 31/08 apocalypse
[/code]

axn 2010-04-17 01:50

[QUOTE=fivemack;212069]It will produce a fair number of relations, but handling those is my concern.[/QUOTE]

How much suboptimal is 31/62? If not too much, it might be better for a distributed effort (half the data transfer).

Also, where is the sweet spot (a 10m-20m range with the highest yield)? It might be good to sieve that region with 16e siever.

schickel 2010-04-17 07:02

Bah! I'd like to help, but no joy:[code]Problem signature:
Problem Event Name: APPCRASH
Application Name: gnfs-lasieve4I15e.exe
Application Version: 0.0.0.0
Application Timestamp: 4aedfe21
Fault Module Name: gnfs-lasieve4I15e.exe
Fault Module Version: 0.0.0.0
Fault Module Timestamp: 4aedfe21
Exception Code: c0000005
Exception Offset: 0000000000011e87
OS Version: 6.0.6002.2.2.0.768.3
Locale ID: 1033
Additional Information 1: 80da
Additional Information 2: 1328344b499912ff3e8828ca4c34a6a9
Additional Information 3: 0d73
Additional Information 4: 84afef7ad7d4a32e4c1f174534df7dd9[/code]System is an AMD X2 running Visa64, 8GB RAM.

I'm pretty sure I grabbed the 64-bit Windows binaries, but I'll double-check after work tomorrow....

axn 2010-04-17 14:16

[QUOTE=schickel;212119]
I'm pretty sure I grabbed the 64-bit Windows binaries, but I'll double-check after work tomorrow....[/QUOTE]

If you're using windows, grab the 32-bit binaries - they're faster.

chris2be8 2010-04-17 16:29

I'd like to help once the work I'm doing has finished. I've got 2 desktops with a 3GHz CPU with hiperthreading and 1 laptop with a 1.6GHz dual core CPU without hiperthreading (but it will run the siever with 64bit asm code which is about 1.8 times faster). But I'm not sure if they have enough RAM.

How much memory does 15i need?

Could you post a sample command line for Linux.

How do you want the output sending to you?

Thanks in advance

Chris K

henryzz 2010-04-18 14:03

[quote=chris2be8;212293]How much less? I want to run two copies on systems with 2Gb so how much does that leave the OS (I can probably stop X windows if necessary)?

Chris K[/quote]
I have successfully done it with 4 cores in the past I am pretty certain(they might have been 14e).

fivemack 2010-04-18 16:21

Hi Chris.

15e, in the early ranges that are available for sieving at the moment, takes about half a gigabyte virtual memory and about a quarter-gigabyte of real memory; you'll have no trouble at all running two copies on a 2G machine, probably no trouble running four on a 2G machine.

If your 3GHz CPU with hyperthreading is an i5 or similar, run one job per hyperthread; if it's a P4 you should probably just run one job.

I'll try to get an FTP server working on fivemack.dyndns.org for uploads.

apocalypse 2010-04-18 16:34

I'll reserve 24M to 28M since it seems to be free.

Also, I think the command line should be
[code]./gnfs-lasieve4I15e -r 2801_79.poly -o <output_file_name> -f <start_of_range> -c <length_of_range>[/code]
[QUOTE=fivemack;212069]
Use the 15e siever, and we'll start by sieving 20 million to 320 million on the [B]rational[/B] (-r) side.[/QUOTE]

jyb 2010-04-18 19:59

[QUOTE=fivemack;212069]Use the 15e siever, and we'll start by sieving 20 million to 320 million on the [b]rational[/b] (-r) side.
[/QUOTE]

I'll be finished with 91M-92M by tonight. What sort of yield are you expecting? I'm seeing around 1.8-1.9 relations/q. In what format should the relations be deposited on your FTP server?

fivemack 2010-04-18 22:32

[QUOTE=jyb;212325]I'll be finished with 91M-92M by tonight. What sort of yield are you expecting? I'm seeing around 1.8-1.9 relations/q. In what format should the relations be deposited on your FTP server?[/QUOTE]

I'm getting 2.05r/q at 20M.

The FTP server is now up; please put relations in the 2801-79 directory with names like 91M-92M.

As for format: plain text, zip, gzip, bzip2 ... basically anything that I can decompress with Linux software. Please don't gratuitously deposit a list of (x,y) pairs, though obviously I [i]could[/i] reconstruct relations from that.

jyb 2010-04-19 03:50

[QUOTE=fivemack;212344]I'm getting 2.05r/q at 20M.

The FTP server is now up; please put relations in the 2801-79 directory with names like 91M-92M.

As for format: plain text, zip, gzip, bzip2 ... basically anything that I can decompress with Linux software. Please don't gratuitously deposit a list of (x,y) pairs, though obviously I [i]could[/i] reconstruct relations from that.[/QUOTE]

91M-92M.bz2 is now there. Final yield was 1.866216 r/q for that range.

BTW, I intended my last message to be my reservation, so I guess the "??" can be 18/04, in case it matters.

chris2be8 2010-04-20 20:53

I didnt get very far:
[chris@localhost o2801_79]$ ../../src/gnfs-lasieve4I15e -r o2801_79.poly -o out28-28.1M -f 28000000 -c 100000
Error: the polynomials don't have a common root:[code]
c0: -1
c1: 0
c2: 0
c3: 0
c4: 0
c5: 0
c6: 2801
Y0: 653236457112135142391905382282254342883556401
Y1: -1[/code]

I can't see any errors in the poly:
[chris@localhost o2801_79]$ cat o2801_79.poly[code]
n: 77728012992649681652257138568390176857212890117660654903605188181803793527234365966882492843633887478337495016140443699327828796305552459390410906211419468091843928832883028959106125700776135225351872298944331308559262827186539091839208835236997045999057406305939986879
skew: 0.266
c6: 2801
c0: -1
Y1: -1
Y0: 653236457112135142391905382282254342883556401
lpbr: 32
lpba: 32
mfbr: 64
mfba: 64
alambda: 2.6
rlambda: 2.6
alim: 134217727
rlim: 134217727[/code]

Any ideas?

Chris K

Batalov 2010-04-20 21:19

you need a newer binary, e.g. from [URL="http://gilchrist.ca/jeff/factoring/nfs_beginners_guide.html"]Jeff's site[/URL].
(the binary you have is fairly old - it appears to have the 255-digit limit which was removed something like a year ago)

Batalov 2010-05-12 19:11

My comp will probably (fingers crossed) be relatively free after tomorrow night, so I'll take a crack at my usual 2% (or is it 'my two cents'?).
I'll take the 67-70M and 134-137M regions (will test the schedule arrays and SCHED_TOL, yada-yada-yada).

apocalypse 2010-05-20 23:39

I've had trouble uploading my latest file, and now I can't reach fivemack.dyndns.org at all. Is something wrong, or am I just unlucky?

[CODE]$ ftp fivemack.dyndns.org
ftp: connect: No route to host

$ ping fivemack.dyndns.org
PING fivemack.dyndns.org (86.173.95.174) 56(84) bytes of data.
^C
--- fivemack.dyndns.org ping statistics ---
36 packets transmitted, 0 received, 100% packet loss, time 35280ms
[/CODE]

fivemack 2010-05-20 23:54

I don't know whether it's my ADSL-to-wifi box, my wifi-to-ethernet bridge or my ethernet switch, but sometimes my network connection falls over until I've rebooted some selection of those devices in some order.

Batalov 2010-05-21 00:30

I got lucky. The transfer was stale for a few minutes and then finished fine (Tom fixed the connection probably, right then).

LatSieveTimes: 0.497Ms like a clock for each of the 67M+1M intervals (times 3), then 0.544Ms for 134M+1M (and less yield, but that's all natural) on PhenomII 940 at 3.25GHz. There were no BAD_SCHEDs or yield dips /[SIZE=1]pats himself on the back[/SIZE]/. Going on for another ~1.1 Megasecs.

Batalov 2010-05-25 19:25

Done with 134-137M. Will upload the last MQ-chunk when possible.
The last timings and counts were:
[FONT=Arial Narrow]134-135M : 0.54Ms : 1738288 unique, 1699 duplicates (removed)[/FONT]
[FONT=Arial Narrow]135-136M : 0.54Ms : 1727485 unique, 1709 duplicates (removed)[/FONT]
[FONT=Arial Narrow]136-137M : 0.54Ms : 1730639 unique, 1683 duplicates (removed)[/FONT]

Ran mock filtering on 6MQ-data; there were ~60K intra-range duplicates and 215 wasted relations with 32-bit+ b values (very few, not a problem).

fivemack 2010-05-26 10:32

The new box forwards packets from adsl to wifi very much more reliably than the old box, does not take two-minute thumb-twiddling breaks every ten minutes, does not cause "YOUR WIFI CONNECTION IS COMPROMISED" messages to appear on my Mac Mini whenever it twiddles its thumbs, and has a nice clear configuration interface.

However, the settings configured in the nice clear packet-forwarding configuration interface don't seem to have any effect on the actual forwarding of actual packets. So fivemack.dyndns.org is down until I get this sorted out, sorry.

mdettweiler 2010-05-26 12:36

[quote=fivemack;216180]The new box forwards packets from adsl to wifi very much more reliably than the old box, does not take two-minute thumb-twiddling breaks every ten minutes, does not cause "YOUR WIFI CONNECTION IS COMPROMISED" messages to appear on my Mac Mini whenever it twiddles its thumbs, and has a nice clear configuration interface.

However, the settings configured in the nice clear packet-forwarding configuration interface don't seem to have any effect on the actual forwarding of actual packets. So fivemack.dyndns.org is down until I get this sorted out, sorry.[/quote]
As a suggestion for the meantime, you guys might want to try uploading relations to [URL]http://www.sendspace.com/[/URL] -- they use that over at the Aliquot project rather extensively for their team sieves now, and it seems to be reliable and decently fast. And I believe files are stored for at least 30 days (and that's reset every time somebody downloads the file).

Granted, it does have the singular disadvantage that fivemack will have to download the results again after they're uploaded (as opposed to having them uploaded directly to him in one fell swoop), but it might still be useful in helping to clear through some of this upload-pending backlog that's built up.

smh 2010-06-01 18:46

[QUOTE=Andi47;216970]In Windows (64 bit, but 32 bit sievers as they are a bit faster in windows) that's ~414 MB per thread.[/QUOTE]Consider running a 64 bit linux virtual machine. The 64 bit linux lasieve is quite a bit faster then the 32 bit windows version.

Andi47 2010-06-05 19:16

[QUOTE=Andi47;217470]Have you got my relations?

BTW: There are lots of reservations which haven't made it into the first posting yet.[/QUOTE]

Adding my relations to the relation counts given in the 1st posting should give a total relation count of [B]95,411,621[/B].

How many relations are needed? (sieving from 20-320M should give less than 392M raw relations (95.4M/(93-20)*(320-20) = ~392.1M) ) [b]fivemack[/b]: your calculation is wrong because not all the 20-93 range has yet been reported ... I'm expecting 500+M raw relations from the 300MQ.

fivemack 2010-06-05 21:37

Please call me an idiot.

The problem was that the dyndns record hasn't been updating, probably because I put 'fivemack' rather than 'fivemack.dyndns.org' in a box on the router configuration, and the router doesn't report whatever error dyndns.org sends it when you give it a wrong domain name. I'd been fussing about with port-forwarding and firewall configurations ...

ftp to fivemack.dyndns.org now appears to work. Sorry about the problem, and I look forward to a tsunami of uploads.

Andi47 2010-06-06 06:03

[QUOTE=Fivemack;217502][b]fivemack[/b]: your calculation is wrong because not all the 20-93 range has yet been reported ... I'm expecting 500+M raw relations from the 300MQ.[/QUOTE]

Ooops, I missed that. (And I almost missed your edit - they are "easier" to miss than a new posting, especially when you click the "show new postings" button).

Recalculating (including axn's new relations) gives:

[code]
Total Q sieved so far: 49.2M
Total relsations so far: 99230661
Relations per Q: 2.017
Rels/Q in the 90M range: ~1.86
[/code]

fivemack 2010-06-06 10:35

One tsunami of relations duly received, thanks.

Andi47 2010-06-06 12:17

1 Attachment(s)
New relation count:

[CODE]
total Q's sieved: 69.2M
total relations: 137,601,215
relations/Q 1.988[/CODE]

I have attached a diagram which shows the rel/Q for each Q range.

chris2be8 2010-06-06 18:45

Reserving 106M-108M. 100M-102M should finish tomorrow.

74M-76M yielded 3855657 relations. But I still can't connect to fivemack.dyndns.org

Chris K

chris2be8 2010-06-06 20:16

I'm still having problems. I can log on, but can't establish a data connection:
[code]
chris@dhcppc63:~/ggnfs/tests/o2801_78> ftp fivemack.dyndns.org
Connected to fivemack.dyndns.org.
220 Welcome to the 2^859-1 temporary upload server
Name (fivemack.dyndns.org:chris): anonymous
331 Please specify the password.
Password:
230 Login successful.
Remote system type is UNIX.
Using binary mode to transfer files.
ftp> ls
229 Entering Extended Passive Mode (|||37602|)

receive aborted. Waiting for remote to finish abort.
ftp> cd 2801-79
250 Directory successfully changed.
ftp> ls
229 Entering Extended Passive Mode (|||59201|)

receive aborted. Waiting for remote to finish abort.
ftp> put 61.2M-61.4M.gz 61.2M-61.4M.e.gz
local: 61.2M-61.4M.gz remote: 61.2M-61.4M.e.gz
229 Entering Extended Passive Mode (|||56167|)
500 Illegal EPRT command.
200 PORT command successful. Consider using PASV.

421 Service not available, remote server timed out. Connection closed
ftp>
[/code]
I've not changed anything on my end.

Chris K

fivemack 2010-06-07 08:41

Chris: does it work better if you say PASV before trying to use other commands? It looks as if your ftp client might be trying to use features which don't get through my new firewall ... I know the ftp server works for some people, it works when I test it from outside and other users have put 20 million relations onto it recently.

chris2be8 2010-06-07 16:51

I think I've found the fix. After reading the man page for ftp I tried epsv:
ftp> epsv
EPSV/EPRT on IPv4 off.
ftp> ls
227 Entering Passive Mode (81,159,163,3,50,206)
150 Here comes the directory listing.
drwxrwxrwx 2 1000 1000 4096 Jun 07 12:53 2801-79
drwxrwxrwx 2 1000 1000 4096 Jan 14 20:22 M887

And I've managed to upload 61.2M-61.4M.f.gz successfully. I'm uploading the next file now.

Chris K

chris2be8 2010-06-08 18:09

100M-102M.gz uploaded. 3673946 relations.

Chris K

richs 2010-06-08 23:21

I was wondering if someone would kindly give me some advice. I am sieving 60M-61M and was about 69% complete when I noticed that the siever had not incremented over the past few hours. I shut down the siever with control-C.

The number in .last_spq0 is 60690713. How do I restart without losing what I have done so far? I already backed up .last_spq0 and 2801_79.out (my output file).

Thanks,

Rich

apocalypse 2010-06-09 00:41

[QUOTE=richs;217862]I was wondering if someone would kindly give me some advice. I am sieving 60M-61M and was about 69% complete when I noticed that the siever had not incremented over the past few hours. I shut down the siever with control-C.

The number in .last_spq0 is 60690713. How do I restart without losing what I have done so far? I already backed up .last_spq0 and 2801_79.out (my output file).

Thanks,

Rich[/QUOTE]

I think this should work.
[CODE]gnfs-lasieve4I15e -r 2801_79.poly -o 2801_79.out -f 60690713 -c 309287[/CODE]

Batalov 2010-06-09 01:15

[quote=richs;217862]I was wondering if someone would kindly give me some advice. I am sieving 60M-61M and was about 69% complete when I noticed that the siever had not incremented over the past few hours. I shut down the siever with control-C.

The number in .last_spq0 is 60690713. How do I restart without losing what I have done so far? I already backed up .last_spq0 and 2801_79.out (my output file).
[/quote]
1. If it had been a BSOD or planned restart, then you could probably repeat the last command line and add -R to the end of it. Backup the output file but leave it in this directory as well (because -R scans the file).
1a. Your binary may not know -R option; then indeed override -f and -c as suggested above; and get a newer one from Jeff's site.
2. However, if the progress report had not incremented (and the computer was not swapping), then it could have been stuck in an infinite loop. What kind of a binary was it? Some binaries are patched for this, old ones are not, but there's another consequence for you - it may get stuck again, on the same value.
3. So, if it starts and gets stuck again, kill it and increase the -f value beyond the "stuck" special_q. Note: the "stuck" q=... on the screen is the previous special_q! The next primes are 60690727 and 60690733. If it get stuck on 60690727, proceed to next, etc. Even old binaries get stuck on one q perl many million, so you will probably not hit another loop.

Tell us later, how it proceeded. -Serge

[COLOR=green]P.S. The 64-bit binary didn't get stuck. Here's the number of relations for some next spq0's:[/COLOR]
[COLOR=green] 29 60690713
30 60690727
33 60690733
45 60690739
40 60690757
8 60690761
27 60690779
43 60690781
37 60690787
37 60690803[/COLOR]

richs 2010-06-09 01:15

Thanks, apocalypse. Looks like I'll have two output files which I assume should not be a problem.

richs 2010-06-09 01:44

The binary was svn374-win32-p4 downloaded mid-May.

Here's what happened since I wasn't sure of the syntax for using -R:

C:\Program Files\GGNFS>gnfs-lasieve4I15e -r 2801_79.poly -o 2801_79.out -f 60690713 -c 309287 -R
Resuming with -f 60999629 -c 371
Warning: lowering FB_bound to 60999628.
^C
C:\Program Files\GGNFS>gnfs-lasieve4I15e -r 2801_79.poly -o 2801_79.out -f 60690713 -c 309287
Warning: lowering FB_bound to 60690712.

I had already made a back-up of the output file. Since the program started at 60999628, I knew that I had the -R syntax incorrect, so I stopped and restarted as shown. I will end up with two output files for this range.

Thanks for the advice since I am just learning how this works.

mdettweiler 2010-06-09 04:34

[quote=richs;217870]The binary was svn374-win32-p4 downloaded mid-May.

Here's what happened since I wasn't sure of the syntax for using -R:

C:\Program Files\GGNFS>gnfs-lasieve4I15e -r 2801_79.poly -o 2801_79.out -f 60690713 -c 309287 -R
Resuming with -f 60999629 -c 371
Warning: lowering FB_bound to 60999628.
^C
C:\Program Files\GGNFS>gnfs-lasieve4I15e -r 2801_79.poly -o 2801_79.out -f 60690713 -c 309287
Warning: lowering FB_bound to 60690712.

I had already made a back-up of the output file. Since the program started at 60999628, I knew that I had the -R syntax incorrect, so I stopped and restarted as shown. I will end up with two output files for this range.

Thanks for the advice since I am just learning how this works.[/quote]
I belive what you need to do for -R is to use the original -f and -c parameters:

gnfs-lasieve4I15e -r 2801_79.poly -o 2801_79.out -f 60000000 -c 1000000 -R

That should then automatically scan the file, pick up where it left off and continue to the end of the range. (Disclaimer: I haven't actually tried this. The last time I used a ggnfs siever standalone, i.e. outside of factmsieve.pl/py, was before the -R feature was implemented and thus I had to use the "old fashioned" resume method that you're using now. Someone else with more experience should be able to confirm whether what I suggested is correct.)

Mini-Geek 2010-06-09 12:39

[QUOTE=mdettweiler;217885]I belive what you need to do for -R is to use the original -f and -c parameters:[/QUOTE]

Yep, this is right. :smile:

Andi47 2010-06-09 13:37

[QUOTE=mdettweiler;217885]I belive what you need to do for -R is to use the original -f and -c parameters:

gnfs-lasieve4I15e -r 2801_79.poly -o 2801_79.out -f 60000000 -c 1000000 -R

That should then automatically scan the file, pick up where it left off and continue to the end of the range. (Disclaimer: I haven't actually tried this. The last time I used a ggnfs siever standalone, i.e. outside of factmsieve.pl/py, was before the -R feature was implemented and thus I had to use the "old fashioned" resume method that you're using now. Someone else with more experience should be able to confirm whether what I suggested is correct.)[/QUOTE]

I have done this quite often and it works like you said. For me it has always worked. (the only times it had failed turned out to be input errors, i.e. a typo in the command line or when I accidently typed "-a" when it should have "-r", or vice versa.)

If the last line of the output file is truncated (that happens quite often), an error message like "Warning: incomplete line. If it's just a few, it's ok, they will be skipped". This is nothing to worry about, as long as you don't get hundreds of these messages.

chris2be8 2010-06-20 19:02

Reserving 137M-139M.

Chris K

PS. This thread needs a post that isn't deleted to move it back up to near the top of the forum.

[SIZE="1"]Let's make it sticky until sieving is finished [/SIZE]

richs 2010-06-21 23:03

60-61M complete and submitted.

Reserving 139-140M.

apocalypse 2010-06-23 04:05

115M-119M completed. 7160853 relations in the range (7130008 unique). uploading now.

Reserving 140M-144M

chris2be8 2010-06-23 16:15

113M-115M uploaded. 3588854 relations.

Chris K

Gimarel 2010-06-24 10:00

Reserving 144M-148M

Andi47 2010-06-26 07:07

1 Attachment(s)
Current relation count:

[CODE]
total Q's sieved 110,200,000
total relations 212,509,920
relations / Q 1.928
[/CODE]

fivemack 2010-06-26 10:12

That big notch at the start is tempting me to resieve the region that I sieved on the wrong side, for purely aesthetic reasons; it would be mathematically inefficient.

Andi47 2010-06-26 10:20

[QUOTE=fivemack;219943]That big notch at the start is tempting me to resieve the region that I sieved on the wrong side, for purely aesthetic reasons; it would be mathematically inefficient.[/QUOTE]

Why would it be inefficient? Is the duplication rate higher when a range is sieved on both sides?

BTW: Some more figures:
[CODE]
max. rel/Q: 2.092 in the range of 24-28M
your notch: 2.000 rel/Q (almost near the maximum)
leading edge: ~1.73 rel/Q (currently around Q~137M)[/CODE]

fivemack 2010-06-26 19:43

For 19.8-112 + 113-137, I have

[code]
Sat Jun 26 12:29:21 2010 found 39584889 hash collisions in 223071581 relations
Sat Jun 26 12:36:44 2010 found 37575436 duplicates and 186716113 unique relations
Sat Jun 26 13:55:49 2010 begin with 186716113 relations and 239612437 unique ideals
[/code]

which sounds as if a huge and ugly matrix might be possible about the time we reach q0=200; will be necessary to sieve a bit beyond that to get a practical matrix.

apocalypse 2010-07-26 15:44

Completed 174M-178M. 6251616 relations in the range, 6232926 unique. Will upload later.

Reserving 202M-206M.

fivemack 2010-07-30 14:02

I'm back in the country and ssh.fivemack.org is accepting relations again.

chris2be8 2010-08-07 19:19

Reserving 216M-220M.

Chris K

apocalypse 2010-08-10 00:42

Completed 212M-216M. 5762377 relations in the range, 5747945 unique.

Could not connect to ftp. Will upload later.
[CODE]$ ftp fivemack.dyndns.org
ftp: connect: Connection timed out
[/CODE]

Reserving 220M-224M.

fivemack 2010-08-10 08:39

Sorry about the ftp outage: my ADSL router got a bit hot and stopped routing. I turned it off overnight and it seems OK now.

chris2be8 2010-08-10 18:00

208M-212M completed and uploaded. 5807956 relations.

Chris K

fivemack 2010-08-11 08:51

OK, we've got enough relations for a matrix, but at the moment the matrix is enormous (27.8M cycles - I'm having to use a 32G machine just to build it). I propose that we keep sieving until the end of August, and then the matrix shouldn't take much longer than the whole of September to run.

Thanks to everyone for staying with me on this rather large computation.

Andi47 2010-08-12 10:34

1 Attachment(s)
[QUOTE=fivemack;224874]OK, we've got enough relations for a matrix, but at the moment the matrix is enormous (27.8M cycles - I'm having to use a 32G machine just to build it). I propose that we keep sieving until the end of August, and then the matrix shouldn't take much longer than the whole of September to run.

Thanks to everyone for staying with me on this rather large computation.[/QUOTE]

If everything goes well, I can upload my range around Sept. 1st or 2nd, is this ok?

P.S.: Current relation count:

[CODE]
sieved Q-range: 194.2M
relations so far: 345,712,217
relations per Q: 1.780
[/CODE]

fivemack 2010-08-12 13:58

That's OK, I'll wait for you. The current matrix is really rather too big (though at least it fits in 12GB to run, even if not to create): estimated runtime is 1500 hours on the i7, which is basically two months.

fivemack 2010-08-16 13:52

Reserving 230-240

chris2be8 2010-08-16 16:35

Reserving 240M-244M.

This will take me until about the end of August, which is probably a good time for me to stop.

Chris K

fivemack 2010-08-27 10:58

230-240 is now finished, I'm doing 15-19.8 at the moment, which ought to be finished in time to pick up andi47's results.

So: [COLOR="Red"]RESERVATIONS CLOSED[/COLOR]

Andi47 2010-08-27 11:21

[QUOTE=fivemack;227254]230-240 is now finished, I'm doing 15-19.8 at the moment, which ought to be finished in time to pick up andi47's results.

So: [COLOR="Red"]RESERVATIONS CLOSED[/COLOR][/QUOTE]

Current relation count: 384.9M

I think I can collect my relations around Tuesday or Wednesday next week and upload them in the evening (GMT+2) on the same day.

jasonp 2010-08-27 14:07

Would it be worthwhile to use a dollop of HPC to solve the matrix?

fivemack 2010-08-27 14:18

If there's a dollop available, and if it would be considered a worthwhile test case (it'll probably be >20M, which might be biggest-one-yet for the parallel LA)

frmky 2010-08-27 20:41

Unfortunately, the Teragrid account's nearly tapped out. I've got one dollup left which I want to save for 5,409-. With luck there will be a fresh supply in a month or so, but I won't know until then. I could run it using MPC (8 computers, GigE connected) though. :smile:

chris2be8 2010-08-28 08:28

240M-244M completed and uploaded. 5488342 relations.

Chris K

Andi47 2010-08-31 06:45

[QUOTE=fivemack;227254]230-240 is now finished, I'm doing 15-19.8 at the moment, which ought to be finished in time to pick up andi47's results.

So: [COLOR="Red"]RESERVATIONS CLOSED[/COLOR][/QUOTE]

It seems that my run has crashed during my holiday:

206-207 is ~25% done
207-208 seems almost done.

I will upload what I have tonight or tomorrow evening.

apocalypse 2010-08-31 08:02

248M-252M Complete. 5419978 relations in the range, 5408212 unique. Uploading now.

Andi47 2010-09-01 20:06

[QUOTE=Andi47;227854]It seems that my run has crashed during my holiday:

206-207 is ~25% done
207-208 seems almost done.

I will upload what I have tonight or tomorrow evening.[/QUOTE]

uploaded:

206-207 (or ~25% of it because it has crashed): [url]http://www.sendspace.com/file/32x18n[/url]
207-208 (should be (almost?) complete): [url]http://www.sendspace.com/file/vn7xbe[/url]

fivemack 2010-09-01 20:22

207-208 appears complete; 206-207 is indeed 25% done.

I'll finish up 206-207 and 19.0-19.8, and start the linear algebra at the weekend.

chris2be8 2010-09-02 16:39

I think we can un-sticky this thread now.

Chris K

fivemack 2010-09-04 07:46

And they're off
 
15-252 is now fully sieved, msieve with target density 100 is poring through 47492098338 bytes of relations (407997768 relations, 318653904 unique)

fivemack 2010-09-04 14:59

[code]Sat Sep 4 14:53:53 2010 matrix is 19514425 x 19514673 (7528.6 MB) with weight 1925523041 (98.67/col)
Sat Sep 4 14:53:53 2010 sparse part has weight 1778442998 (91.13/col)
Sat Sep 4 14:59:46 2010 linear algebra at 0.0%, ETA 865h14m

linear algebra completed 22997 of 19514673 dimensions (0.1%, ETA 869h28m)

10131 nfsslave 20 0 8496m 8.3g 1048 R 343 70.3 609:04.83 msieve[/code]

Should be done in about five weeks

fivemack 2010-10-07 09:10

Unfortunately, the timing has not worked out perfectly: the linalg would finish on Saturday afternoon, but I am going to China for two weeks on Saturday morning and will turn the computers off. So expect factors on about 26 October.

fivemack 2010-10-27 21:27

[code]Wed Oct 27 12:00:11 2010 prp85 factor: 8022468609453611665082019221530953944163549137920344510245747854780835146671963776777
Wed Oct 27 12:00:11 2010 prp184 factor: 9688789919483838708810358343444188308059016146046561072836401903385375007156741636000015490808770627835181274383093018972359887003518258190829470483941361347766634033284584062134510727

(from the fifth dependency, at eight hours per sqrt)[/code]

About 835 wall-time hours on four cores of i7 to do the linear algebra.

Batalov 2010-10-27 22:31

Congrats! That was some heavy-lifting!

debrouxl 2010-10-28 08:05

Great, congratulations :smile:

wblipp 2010-11-06 11:36

:wblipp:

Wow! This was a BIG project that I had not expected to see this year. Oddperfect.org had been advancing ECM on this most-limiting obstruction, but mostly in hope of finding a factor. The base 2801 isn't of interest outside of the OPN community, and the number strains resources. But in April Tom grabbed the bull by the horns and said "let's do it!" Nearly six months later, with huge contributions by dozens of people, it's finished.

Thanks, Tom, and all the other contributors. There are ways to push OPN proofs higher without specific factorizations, but the result is always a higher limit that is constrained by the same missing factorization. This factorization would have determined the limitation of all OPN proofs - all OPN researchers rejoice at this factorization.



:wacky:


All times are UTC. The time now is 03:55.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.