Team sieve #17: c163 from 4788:2509
 

[B]Postprocessing will be done by Joshua2[/B]

[B]Linear algebra is complete! Factors as [URL="http://factorization.ath.cx/search.php?id=127242325"]p61[/URL]*[URL="http://factorization.ath.cx/search.php?id=127242326"]p103[/URL][/B]

[B]Poly:[/B][code]n: 8519618226197093878278613851429334276607578086560582046132272468491047228903383713809367142607155237315475444625223025926760239272565192820910356755196796601078683
# norm 4.832139e-16 alpha -8.365564 e 6.936e-13
skew: 82092126.77
c0: -4506332652721281857439312883465578850580235
c1: 1546557538158535979212952557252545273
c2: -493361789599381258970904335
c3: -1674326900805118095209
c4: -406958861894
c5: 22800
Y0: -51819960992313118269272571928496
Y1: 813459549425076319
rlim: 50000000
alim: 50000000
lpbr: 30
lpba: 30
mfbr: 60
mfba: 60
rlambda: 2.6
alambda: 2.6[/code]Use siever [B]14e[/B], sieve 20M-120M Q.

[B]Command line:[/B][code]gnfs-lasieve4I14e -a 4788_2509.poly -o <output_file_name> -f <start_of_range> -c <length_of_range>[/code] where 4788_2509.poly is the poly file above.

[B]Reservations:[/B][code][B] Range Who[/B]
*19.9M- 20M Andi47 (15e siever; done, 243426 relations)
* 20M- 21M Andi47 (done, 1146336 relations)
* 21M- 22M Greebley (done, 1149732 relations)
* 22M- 40M bsquared (done, 21936082 relations)
* 40M- 41M EdH (done, 1237326 relations)
* 41M- 50M FactorEyes (done, 11286234 relations)
* 50M- 60M fivemack (done, 12196919 relations)
* 60M- 67M fivemack (done, 8146713 relations)
* 67M- 70M Andi47 (done, 3419791 relations)
* 70M- 71M Andi47 (done, 1122887 relations)
* 71M-110M RSALS grid (done, about 33606932 relations)
* 110M-111M Andi47 (done, 964336 relations)
* 111M-115M bsquared (done, 3845345 relations)
* 115M-119M bsquared (done, 3811160 relations)
* 119.0-.4M Andi47 (done, 377383 relations)[/code][B]Total relations received: 104490602[/B]

Use a site such as [URL="http://www.sendspace.com"]Sendspace[/URL] to upload relations.
Please do not use Rapidshare if possible as it has time consuming restrictions.

[QUOTE=10metreh;201619]
Use siever [b]14e[/b], sieve 20M-120M Q.
[/QUOTE]

Have you tested the 15e siever? We might be near the crossover line...

BTW: Reserving 20M - 21M

I will grab 21-22 M

[QUOTE]Postprocessing will be done by ? (4GB RAM needed)[/QUOTE]
Maybe only 2.5GB for the linear algebra, or less with sufficient over-sieving. What about the filtering stage? Tom, how much did the filtering need for the c161 in December?

[QUOTE=Andi47;201631]Have you tested the 15e siever? We might be near the crossover line...[/QUOTE]

I already did.

Here are the tests with [b]14e[/b] (to get the range of Q needed):
[code]Warning: lowering FB_bound to 9999999.
total yield: 837, q=10001009 (0.15551 sec/rel)
Warning: lowering FB_bound to 14999999.
total yield: 1323, q=15001001 (0.13859 sec/rel)
Warning: lowering FB_bound to 19999999.
[color=red][b]total yield: 1107, q=20001001 (0.13996 sec/rel)[/b][/color]
Warning: lowering FB_bound to 29999999.
total yield: 1210, q=30001003 (0.13541 sec/rel)
Warning: lowering FB_bound to 39999999.
total yield: 1262, q=40001021 (0.13996 sec/rel)
[color=red][b]total yield: 1205, q=50001037 (0.13965 sec/rel)[/b][/color]
total yield: 968, q=60001021 (0.15442 sec/rel)
total yield: 706, q=70001011 (0.14997 sec/rel)
[color=red][b]total yield: 1010, q=80001017 (0.15086 sec/rel)[/b][/color]
total yield: 1230, q=90001003 (0.15591 sec/rel)
total yield: 961, q=100001029 (0.16170 sec/rel)
total yield: 1222, q=110001053 (0.16511 sec/rel)
[/code]

And spot-checking with [b]15e[/b] for comparison
[code]Warning: lowering FB_bound to 19999999.
[color=red][b]total yield: 2399, q=20001001 (0.17222 sec/rel)
total yield: 2598, q=50001037 (0.15870 sec/rel)
total yield: 2108, q=80001017 (0.17734 sec/rel)[/b][/color][/code]

Quite a bit slower with 15e.

I could do postproc with my quad core q9600 @ 3ghz w/ 8gb mem unless someone has better. You would have to walk me through because I've never done it except when everything is automated.

[QUOTE=Joshua2;201655]I could do postproc with my quad core q9600 @ 3ghz w/ 8gb mem unless someone has better. You would have to walk me through because I've never done it except when everything is automated.[/QUOTE]

Using msieve, this is easy. You may not be familiar with jobs of this size though, so just so you know what you're getting into:
* expect to not use that PC for several days to a week, as any other process running on it will be slowed to a crawl
* unless you have ECC memory, plan on checking up on it occasionally and restarting from checkpoints as necessary
* unless you also plan on doing a significant amount of sieving, be ready to download many gigabytes of relation data

Still want to? :smile:

Definitely not trying to discourage you, and apologies if you are aware of this already...

[B]reserving 22-40M[/B]

[QUOTE=Joshua2;201655]I could do postproc with my quad core q9600 @ 3ghz w/ 8gb mem unless someone has better. You would have to walk me through because I've never done it except when everything is automated.[/QUOTE]

Do you have at least 20GB of unused disk space? The relations file will be at least 10GB unpacked, and msieve will create some other large-ish files during the processing.

Jason's msieve is easy enough to use, you should not have any difficulty learning it. Basically all you'd need to do is to download and keep track of the relation files produced here, untar/zip/whatever and combine them all into a [b]msieve.dat[/b] file,

make a [b]worktodo.ini[/b] file containing:
[code]
N 8519618226197093878278613851429334276607578086560582046132272468491047228903383713809367142607155237315475444625223025926760239272565192820910356755196796601078683
[/code]

a [b]msieve.fb[/b] file containing the poly in msieve format:
[code]
N 8519618226197093878278613851429334276607578086560582046132272468491047228903383713809367142607155237315475444625223025926760239272565192820910356755196796601078683
SKEW 82092126.77
A0 -4506332652721281857439312883465578850580235
A1 1546557538158535979212952557252545273
A2 -493361789599381258970904335
A3 -1674326900805118095209
A4 -406958861894
A5 22800
R0 -51819960992313118269272571928496
R1 813459549425076319
[/code]

and run:
msieve -v -nc -t 4
(for 4 threads)

Should take less than a week of wall-time, assuming the computer is otherwise relatively idle.

[QUOTE=bsquared;201658]Using msieve, this is easy. You may not be familiar with jobs of this size though, so just so you know what you're getting into:[/QUOTE]
Thanks. I have plenty of diskspace and bandwith, but no ecc, but I could check on it if I know what to look for, but I have be able to use the computer. I always have stuff running and just put it at low priority, that won't work for this? I would think it saying 4gb needed I would be able to do stuff occasionally with 8gb.

You could run msieve with lowest priority, and then you can use your cpu as normal. However I've never been completely satisfied with cpu schedulers in this regard... it always seems like the system is less responsive even if the heavy job is low priority (windows, anyway).

Also, if normal operation for you is to be continuously doing other cpu intensive tasks on multiple cores, then the matrix runtime will be significantly extended. Impatient people around here may object to that. Otherwise I think normal part-time use would be fine, provided you can live with the performance of the system with msieve running in the background.

I'm not really sure how crucial ECC memory is to large matrix jobs; Batalov I know has run into problems before, but I haven't personally. From what I gather, the Lanczos routine will abort with an error message and you need to manually restart it using -ncr.

Isn't the right comparison with 15e using 29-bit large primes?

I normally run cpu intensive stuff in the background at idle, but I would cancel them during the duration of this project. Sometimes I game though or just surf the web.