2^821-1

[mdettweiler]

It looks like there's an error in the reservations table: I reserved 45-45.5 on the R side, not the A side. (However, it really doesn't make much difference to me which side I search, so if it would make things easier I'll just as easily do the A side instead. )

fivemack: sorry, fixed the reservations table

[smh]

I'll reserve 46-47 A+R

[smh]

I'll continue 47 - 48 A+R and will see how far i get.

[fivemack]

Just a reminder that sieving should be finished by tomorrow evening my time (IE about 36 hours from when I post this). We ought to have masses of relations and a nice small matrix, with hope for factors by the start of July.

[mdettweiler]

45M-45.5M R complete, relations uploading as I type.

[Andi47]

I just started the upload of my relations.

[smh]

Just wondering, did you manage to build a matrix, and what's the ETA?

[fivemack]

Originally Posted by smh

Just wondering, did you manage to build a matrix, and what's the ETA?

Yes, I've got a decent matrix (just over 10M on a side), ETA is evening of 24 June. The first half of the oversieving saved significantly more time than it took; the second half perhaps not so much.

The matrix fits very happily on the 4G quad-core:

nfsslave2@sheep:~$ ps -F 14087
UID        PID  PPID  C    SZ   RSS PSR STIME TTY      STAT   TIME CMD
1000     14087 14046 99 886339 3512856 1 18:20 pts/1   Rl    10:53 ../msieve-1.34/msieve -v -ncr -t 4

[fivemack]

The cofactor of 2^821-1 splits as

25348330589061322436059499894718930577798031880081867566547223 * 41866449065688547732334670918138300495432254381696818748989283844419920572286356810694501570678523880104345359476534651949605873238987445628114071

The smaller factor is a P62; the 10129794 x 10130042 matrix (2798.1 MB) with weight 690519820 (68.17/col) sparse part has weight 632200045 (62.41/col) took 374:36:58 to run on a Q6600 quad-core with 4G memory, and the factors were found on the first dependency after seven hours.

Sieving effort was Q=40M to 60M and 70M to 100M, both algebraic and rational sides; from my records it was about 220 CPU-hours per million Q per side, so just under three core2/2400-years for the whole job.

[R.D. Silverman]

Originally Posted by fivemack

The cofactor of 2^821-1 splits as 25348330589061322436059499894718930577798031880081867566547223 * 41866449065688547732334670918138300495432254381696818748989283844419920572286356810694501570678523880104345359476534651949605873238987445628114071

The smaller factor is a P62; the 10129794 x 10130042 matrix (2798.1 MB) with weight 690519820 (68.17/col) sparse part has weight 632200045 (62.41/col) took 374:36:58 to run on a Q6600 quad-core with 4G memory, and the factors were found on the first dependency after seven hours.

Sieving effort was Q=40M to 60M and 70M to 100M, both algebraic and rational sides; from my records it was about 220 CPU-hours per million Q per side, so just under three core2/2400-years for the whole job.

Nice work.

Allow me to ask: What were the factor base bounds? What was
the size of the sieve region per Q?
Also, I presume the 220 hrs/10^6 Q represents data from multiple
machines? I am currently doing 2,1538M and one of my machines processes
a single q in about 10.3 seconds. --> 2860 hours per million q, not 220.

[fivemack]

Originally Posted by R.D. Silverman

Nice work.

Allow me to ask: What were the factor base bounds? What was
the size of the sieve region per Q?

Factor base bounds are min(8e7, Qrat) on the rational side, min(10e7, Qalg) on the algebraic side. Sieve region is 2^15 * 2^14.

Also, I presume the 220 hrs/10^6 Q represents data from multiple
machines? I am currently doing 2,1538M and one of my machines processes
a single q in about 10.3 seconds. --> 2860 hours per million q, not 220.

Ah, I say '10^6 Q' to mean 'Q between N and N+10^6', which corresponds to about 55,000 usable Q values. In the 220 hours, that means about 14.4 seconds per ideal.