Estimated relations Factmsieve

[wombatman]

Alternatively, post the file that ends in (probably) ".resume". Easiest way will be to zip it so you can attach it here. That's what gets messed up when you change the number of threads mid-sieve. I learned that the hard way.

It can be fixed so it's not running duplicate ranges.

[VBCurtis]

I concur- changing the number of threads without manually altering the resume file confuses factmsieve, because it split ranges into 64 pieces but then only 8 of them were completed.
It's easier to let it finish than to hope you can deduce what ranges were skipped *and* the magic invocation to fill those ranges in a way faster than just letting the script limp to the finish line.
The "finish line" is not definite; the script sets a target number of relations somewhat lower than what usually works, as the cost to try to build a matrix a couple of times is generally lower than accidentally sieving way too long (and it's not obvious ahead of time exactly how many relations are needed, because an unpredictable number of duplicate relations are generated).
RichD observed that you have an unusual number of duplicate relations; the process naturally duplicates around 10-20% of the relations it finds, but can only make use of unduplicated relations. So, in a sense, 35% of the list you've found is discarded, and the process must run until it gets enough unduplicated relations. The thread-change likely produced the extra duplicates as a side effect; again, the only harm is a small delay.
A C130 on a desktop i7 would normally complete in 15-20 hrs; maybe yours will take 24 or so. Once a matrix builds, it should finish in an hour or so.

[cimpresovec]

Thanks for all the info. I'll let it continue tomorrow then, I don't want to fiddle with it anymore, might break it even more. I hope it will be able to finish the process.

[RichD]

Once you get a matrix built, it will provide a running ETA for the linear algebra phase.

After that, it is only minutes to complete the SQRT phase. Hopefully the solution will be in the first few dependencies. Good luck!

[cimpresovec]

I'm now at almost 16M unique relations but it still doesn't go into the final step. It's at 128.6% of the initial estimate. What is strange is that the log always says

filtering wants 1000000 more relations

It doesn't change. Can there also be a problem here?

[chris2be8]

msieve always says "filtering wants 1000000 more relations" whenever is doesn't have enough relations. Ignore that.

A better clue is:

Sat Jan 16 17:08:20 2016  reduce to 3472844 relations and 4012010 ideals in 32 passes

That suggests you need 10-20% more relations. Once it says more relations than ideals you are probably within 10% of enough.

Chris

[Batalov]

Originally Posted by cimpresovec

Thanks for the help. I'll wait if someone else can advise.

Check if your disk is full, btw.

[cimpresovec]

I just got to the point that I have the matrix. But it throws an error. Since I'm on Windows it can't find the /home/math/mpihost program. Here is the full error.

-> msieve -s .\.\example\example.dat -l .\.\example\example.log -i .\.\example\example.ini -nf .\.\example\example.fb -t
 8 -nc1
read 10M relations
read 20M relations
-> Running matrix solving step ...
-> mpirun -n 3 -hostfile /home/math/mpihosts ./msieve -s .\.\example\example.dat -l .\.\example\example.log -i .\.\examp
le\example.ini -nf .\.\example\example.fb -t 2 -nc2 3,1
Traceback (most recent call last):
  File ".\factmsievem.py", line 2110, in <module>
    ret = run_msieve('-t {0:d} -nc2'.format(LA_THREADS))
  File ".\factmsievem.py", line 398, in run_msieve
    ret = run_exe(MPI_RUN, args)
  File ".\factmsievem.py", line 357, in run_exe
    p = subprocess.Popen([ex] + args.split(' '), **al)
  File "C:\Python27\lib\subprocess.py", line 710, in __init__
    errread, errwrite)
  File "C:\Python27\lib\subprocess.py", line 958, in _execute_child
    startupinfo)
WindowsError: [Error 2] The system cannot find the file specified

What is this mpihosts thing? What can I do?

EDIT 1:
It seams I have the script from some weird source. I tried to run just the last ./msieve part of the command above, but it crashes. So now I'm kinda stuck, if I try to use a different version of the factmsieve script, it deletes the data.

EDIT 2:
I have played around with it a bit, but can't get it to run, msieve gets some error that it can't find some relations. I have added the log.

EDIT 3:
By getting an recent "official" version of the script and with some tinkering, and copy pasting, I got the new script to use the old data and rerun the matrix building part. Now msieve crashes at another part.

-> msieve -s .\.\example\example.dat -l .\.\example\example.log -i .\.\example\example.ini -nf .\.\example\example.fb -t 4 -nc2
linear algebra completed 254 of 1887933 dimensions (0.0%, ETA 2h 3m)
Return value -1073740940. Terminating...

Here is the updated log. Now I really don't know anymore.

[cimpresovec]

New post since I can't edit anymore.
I searched around and found a 64bit version of msieve from wombatman and tried again after getting it to run. It's now past the previous error step and working. I'll run this through tomorrow, it's already too late.

[cimpresovec]

I have managed to factor the number and have reported them to factordb. The final report says 60 hours, but that's probably wrong. Thanks for the help everyone. I'll now try to use this to decrypt the files, hopefully it will work.

[wombatman]

Congrats! Glad it worked out in the end.