[QUOTE=Batalov;178822]
Do lots of ECM first, of course.[/QUOTE]

I have queued a bunch of curves @ B1=11M

So what is the command to run ecm on this? I have been letting aiqueit handle it up to this point. Might as well learn how to do it by hand and help out at the same time.

[QUOTE=Greebley;178826]So what is the command to run ecm on this? I have been letting aiqueit handle it up to this point. Might as well learn how to do it by hand and help out at the same time.[/QUOTE]

create a file which contains the c157, followed by an endline character. (e.g. named alq4788.2422.txt)

then in the command line type:

[code]ecm -nn -c <number of curves you want to do> <alq4788.2422.txt <B1> >>outputfile.out[/code]

so an example command line might look like this:

ecm -nn -c 100 <alq4788.2422.txt 11e6 >>alq4788.2422_11e6.out

the -nn option makes ecm run in lowest priority mode (so that it doesn't interfer with other programs like word or excel), -c specifies the number of curves you want to run, B1 is the "B1" bound (see the ECM article in the mersennewiki). For this job, use either 11e6 or 43e6 as B1 bound.

[quote=Greebley;178826]So what is the command to run ecm on this? I have been letting aiqueit handle it up to this point. Might as well learn how to do it by hand and help out at the same time.[/quote]
Here's an example command:
[code]ecm -c 5 -one 43000000 < c157.txt[/code]ecm: the name of your ECM.exe executable (.exe part is optional on Windows, of course)
-c 5: Says to run 5 curves. Adjust accordingly.
-one: Stops if it finds a factor.
43000000: the B1 value. Adjust accordingly.
(optional after the B1 value is the B2 value)
" < c157.txt" sends ECM the data in c157.txt, so before you run this command make a c157.txt with the c157 in it. (without any "n: " before it or anything)

p.s. at this size and B1, it took me (one core of an Athlon 2.5 GHz dual core) just over 10 minutes for step 1 and just under 4 minutes for step 2, so if you don't see any "step 1 completed" for a few minutes, don't think it froze or anything. :smile:

echo number | ecm -c count -one b1limit | tee -a ecmout

where "number" is the number to factor, "count" is how many curves you want to do, and "b1limit" is the size of B1 you want. The -one is so that the program will quit when it finds a factor even if the cofactor is still composite (since at that point it will be easier to GNFS than continuing ECM).

You definitely want to output the results to a file so you won't lose the results if the system goes down. I like using tee for this. In this example the output will be appended to a file named ecmout.

So if you want to do 100 B1=43M curves on this number you would do

[code]echo 2179298864647941593089329843480960377234329034733632473654418628122413680526625181566302348541331686710338443591873700839852521892809407336286970457974236409 | ecm -c 100 -one 43e6 | tee -a ecmout[/code]

FYI I just re-checked the sequence from the database, since others have mentioned errors with other sequences. 4788 is right.

[QUOTE=jrk;178843]FYI I just re-checked the sequence from the database, since others have mentioned errors with other sequences. 4788 is right.[/QUOTE]

this should be done with [b]all[/b] sequences in the database, because there are seqs with false computation of an index and if such incorrectness will be continued with the workers, they won't find those errors, only with aliqueit or other checking progs!

see the thread with error-seqs!

3000@1e7 complete, no factors

3200 @ 43M completed last night, no factors.

Are the P-1 and P+1 sufficient? The DB lists P-1 B1=1000000000 (1e9 or 1G) and B2=205702371522480 (just over 2e14 or 205T) and two runs of P+1 with the same parameters. If not, what bounds would be good, about how long will it take, and how much memory will it use in stage 2? I'd be interested in doing more P-1 and/or P+1 if it would be useful and fit in 1GB of available memory.
From a few short tests of whether GMP-ECM errors or not with the given maxmem and B1, it looks like I can do up to B1=4e9 and would take between 400 and 450 MB. Anyone know if that's remotely right?

1800@43e6 set going over the weekend; by Monday either we'll have a factor or should start polynomial search.