[QUOTE=bsquared;241627]I'm not sure what happens if the curve is aborted abnormally, such as a forced computer shutdown. maybe the -chkpnt flag would be of some use in that case.[/QUOTE]

Just ran a test. If the curve is aborted (at least with a cntl-C signal), the highest bound processed is stored. With a cntl-c somewhere in the middle of a run to 1e6, I got this savefile:

[CODE] % cat savefile.dat
METHOD=ECM; SIGMA=612573498; [B]B1=504349[/B]; N=1553303516292478972434431257669685410562962830016991957242492897198835922479639127670719679; X=0x4da67665395859a7ceabbc8d7a5e5d365fd6e02009fc030313c59f2e3a449c36faa4e5ff932; CHECKSUM=519957310; PROGRAM=GMP-ECM 6.3; WHO=buhrow@****; TIME=Mon Dec 13 10:15:34 2010;
[/CODE]

So it seems pretty robust :)

[QUOTE=bsquared;241629]Just ran a test. If the curve is aborted (at least with a cntl-C signal), the highest bound processed is stored. With a cntl-c somewhere in the middle of a run to 1e6, I got this savefile:

[CODE] % cat savefile.dat
METHOD=ECM; SIGMA=612573498; [B]B1=504349[/B]; N=1553303516292478972434431257669685410562962830016991957242492897198835922479639127670719679; X=0x4da67665395859a7ceabbc8d7a5e5d365fd6e02009fc030313c59f2e3a449c36faa4e5ff932; CHECKSUM=519957310; PROGRAM=GMP-ECM 6.3; WHO=buhrow@****; TIME=Mon Dec 13 10:15:34 2010;
[/CODE]

So it seems pretty robust :)[/QUOTE]

Much obliged. Is there a 64 bit Windows binary available?
(i.e. that takes advantage of the 64x64 multiplies)?

I have a 32-bit binary.

I just don't have the resources to do any of the remaining Cunningham
composites via NFS. I will shortly have to turn to something else.

[QUOTE=R.D. Silverman;241632]Much obliged. Is there a 64 bit Windows binary available?
(i.e. that takes advantage of the 64x64 multiplies)?
[/QUOTE]

Yep, here is one: [url]http://www.mersenneforum.org/showpost.php?p=230017&postcount=204[/url].

There are probably others elsewhere in that thread for different platforms, if needed. Or I could build you one, just let me know.

[QUOTE=R.D. Silverman;241602]
I may also do the remaining homogeneous numbers with exponent under
200.[/QUOTE]

All those guys have 3000 curves @ B1=43e6 - around 40% of t50 - from me.

3^499+2^499, which is the only other remaining number from the previous batch, has had 2*t50 @ B1=43e6.

I may have put a bit more into some of them, but I'll stick with conservative totals, so as to minimize the chance of an ECM miss.

[QUOTE=R.D. Silverman;241602]I may also do the remaining homogeneous numbers with exponent under 200. Another possibility is to run SNFS on some of the smaller unfactored Fibonacci/Lucas numbers.[/QUOTE]A deliberately mischevous suggestion is that you may wish to contribute to a project of mine where there are some relatively easy candidates. The Cullen and Woodall numbers are presently at SNFS-difficulty in the high 600 digits, GNFS difficulty in the mid-500 bits and ECM at around the mid 40-digit level. Their generalized counterparts are significantly easier.

PM or email if you may be interested.

Paul

[QUOTE=R.D. Silverman;241632]Is there a 64 bit Windows binary available?
(i.e. that takes advantage of the 64x64 multiplies)?[/QUOTE]
Jeff Gilchrist maintains a nice assortment [URL="http://gilchrist.ca/jeff/factoring/index.html"]here[/URL].

[QUOTE=R.D. Silverman;241602]My work on 2,1870L is about 2/3 done.

My company shuts down from 12/24 until 1/3, so I won't get much done
during that time.

I am undecided about what I will do next. I may go after the 2+ table
with high ECM bounds, but with only ~20 machines do not have much hope
of accomplishing anything. [/QUOTE]

In case that ECM after 2+ tables go unproductive, then
would you mind doing - clearing out with all that easy numbers from that base 3 extension table candidates?

Can you be able to do with that, or rather are you interested in that 3,605+ right now which is in main tables as of now? Ok, it may be the case that Linear Algebra for this number may not fit upon your machine, although you may be able to do so with sieving for this number.

Christmas factors
 

I have factored 11^203-10^203 with SNFS:

[CODE]prp68 factor: 39335529070002522184476596661339822469180852209326204156225499170017
prp74 factor: 13887701777825996803419001648448604886605761981008503109397032671761985263[/CODE]

Mail to Paul is already sent.

Merry Christmas!

11^238+4^238: sextic or octic?
 

I am considering to SNFS 11^238+4^238. With the obvious sextic poly, the difficulty is 212.

What is the difficulty with an octic? is it [SUB]10[/SUB]log((11^238+4^238)/(11^14+4^14))=233.3?

c8: 1
c7: -1
c6: -7
c5: 6
c4: 15
c3: -10
c2: -10
c1: 4
c0: 1
Y1: 101938319743841411792896
Y0: -144209936106571291631713992017

Edit: I will do it with the sextic, but have I calculated the "octic" difficulty correctly?

Edit2: queuing a whole lot of ECM curves first. (is ~70% of t50 approx. right? How many curves did this number have?)

Looks correct. So it is a double whammy: octic [I]and[/I] harder.
(Octics cannot compete with a sextic in this range.)

[SIZE=1]Incidentally, I've looked at some b[SUP]17n[/SUP]+-1 in Cunninghams (e.g. 7,323+-): no go as octics.[/SIZE]

[QUOTE=Andi47;243376]I am considering to SNFS 11^238+4^238.

[snip]

queuing a whole lot of ECM curves first. (is ~70% of t50 approx. right? How many curves did this number have?)[/QUOTE]

I calculate that it needs about 3300 curves with B1 = 43e6. It's received about 110 via Paul's ECMnet server. No idea if anyone else has thrown more curves at it.