Status

[R.D. Silverman]

What's the status of 5,317-, 6,283-, 5,323- etc? I understand that
they finished sieving some time ago.

[Wacky]

Paul has been processing 6,283-.
On Sept 5, he reported that he had produced a 4.75M matrix and that it would require about 12 more days to run the Block Lanzcos phase on his machine. So I expect to hear something from him soon.

Subsequently, I began processing 5,317-. I managed to produce a 4.38M matrix but my 32 bit processing is much slower than Paul's 64 bit implementation. I'd done about 20%, but do not expect to finish before late October.

5,323- is our current sieving. We have done about 40% and should complete that sieving mid-October if we are able to continue at the current rate.

[xilman]

Originally Posted by Wacky

Paul has been processing 6,283-.
On Sept 5, he reported that he had produced a 4.75M matrix and that it would require about 12 more days to run the Block Lanzcos phase on his machine. So I expect to hear something from him soon.

I'm typing on a laptop in a hotel room just off Harvard Square right now and accessing my home systems is possible but not entirely simple. When I last checked, some hours ago, the matrix still had a few hours to run. Therefore, it should be close to finished now.

If I can complete the factorization from here I will. Otherwise, it will have to wait until I return home (doubtless jetlagged) on Thursday.


Paul

P.S. Actually, it is impossible right now. It appears that my home ADSL link is down.

[xilman]

Originally Posted by xilman

I'm typing on a laptop in a hotel room just off Harvard Square right now and accessing my home systems is possible but not entirely simple. When I last checked, some hours ago, the matrix still had a few hours to run. Therefore, it should be close to finished now.

If I can complete the factorization from here I will. Otherwise, it will have to wait until I return home (doubtless jetlagged) on Thursday.


Paul

P.S. Actually, it is impossible right now. It appears that my home ADSL link is down.

The linear algebra had finished by this morning (Boston time) and so I kicked off the square root stage. The latter is expected to take an hour or more so I left it running. After the day's business here I tried to pick up the factors but the ADSL was down again.

With a bit of luck I may be able to get the factors tomorrow.


Paul

[Wacky]

Last night, Paul reported to me that he had returned to England and retrieved the following information before going to bed after an exhausting trip.

6,283-

Probable prime factor 1 has 99 digits:
447124831877025366689793129436873423163216894462385022533572097231958503707190221729170132612703957
Probable prime factor 2 has 75 digits:
138457361320915478919381975760508114488979126852819238404548238145324558533

so not an ECM miss, even by Bruce's standards.

I'm sure that it is likely that he will add some more details after he recovers.


On 5,317-, I have completed 30% of the Block Lanczos iterations.

And we have completed approximately 50% of the sieving for 5,323-.

[Jwb52z]

Is the actual NFSNet project website going to be updated soon with all this new information about what has been completed since April? That was the date of the last bit of news on the website. I understand if the person who does that needs to recover from his travelling, but as I help keep one of the biggest DC project websites updated, it would be nice to make some official news about it from the website, no hurry though.

[bdodson]

Originally Posted by Jwb52z

Is the actual NFSNet project website going to be updated soon ... I understand if the person who does that needs to recover from his travelling, ...

Xilman does post-processing after sieving, and sets sieving regions
for the new numbers; not web page maintenance. The person that
used to do the web page and stats retired from NFSNET, and hasn't
been replaced. Fivemack/Tom did some interim updates, perhaps he
could do another round? -bdodson

[Jwb52z]

Thank you so much for the quick answer to my question. There are a great many people who visit the DC site I help with and I'm sure they are interested in what's going on with things. :)

[Wacky]

I am pleased to report that, with the assistance of "frmky" Greg Childers (CalState-Fullerton) and utilizing code from msieve ("jasonp" Jason Papadopoulos), NFSNet has completed the factorization of 5,317-.

Greg reported

Fri Sep 28 12:06:02 2007 prp85 factor: 1173266048118996938584719882501239841331337879112270918586790280760729499132694039331
Fri Sep 28 12:06:02 2007 prp110 factor: 78784317656768133239109671345422644991678073397834197717116145126532590520938482143011654153492533979880370291

Greg

Thanks to all who helped in this effort. I'm sure that Greg will have something to add about his part of the effort.

In the meantime, we continue to sieve for 5,323-, having reached about 2/3 of the estimated interval. With both Paul and Greg capable of doing the post processing in a reasonable timeframe, we can use some additional help in the sieving.

Please join us.

On the behalf of the entire NFSNet Factoring Group,
Richard "Wacky" Wackerbarth

[frmky]

Just a few details... As far as I am aware, this the largest SNFS factorization to date completed with Jason's msieve. Approximately 52 million relations were converted from CWI to GGNFS format, then fed to msieve running on a 1.8 GHz single-core Opteron 144 system with 3 GB of memory in 64-bit Linux.

The filtering took about 4 hours to complete, and used 2.6 GB of memory. It produced a final 4945096 x 4945296 matrix with weight 430746398. After setting aside the dense rows due to the quadratic characters and small primes, Block Lanczos was started on a 4945048 x 4945296 matrix with weight 325847903 (a bit less than 66 nonzero entries per column). Although the calendar time was a bit longer due to concurrently running processes, the runtime was just over 6 days, and used 1.4 GB of memory. Each square root run took about 2 1/4 hours and used about 1 GB of memory. The factors were found on the 5th dependency.

Greg

[jasonp]

Originally Posted by frmky

Just a few details... As far as I am aware, this the largest SNFS factorization to date completed with Jason's msieve.

I must say, this announcement has just made my evening. Richard, is there any information about the matrix generated by the CWI suite, and how long the linear algebra would have taken?