Feedback for new MPQS utility sought

[jasonp]

Originally Posted by roger

I tried the -r command, but unfortunately the program still gives the same result. The output is that there are errors reading the first 5400+ relations. Perhaps Msieve found 5400 relations on the end rather than the corrupted ones at the beginning?
Is there a way around this, such as restarting the composite with a different savefile, then combining them later so that they can Finally give a result?

You basically have to get msieve to collect enough relations so that the 5400 corrupt ones can be ignored but there are still enough good relations left for the postprocessing to start. You can run with -r again if it will help. Alternately,

- save off the old save file
- run msieve from scratch, stopping the run after it claims to have found 6000 relations (i.e. it says 'combined from 6000 partial')
- combine the two savefiles: 'type msieve.dat >> msieve_old.dat', where msieve_old.dat is the old savefile
- restart from the modified savefile

If 6000 relations are not enough, repeat the process until it stops giving errors and begins the postprocessing.

jasonp

[roger]

Thank you, JasonP

That will take so little time compared to beginning over a Third time.

Thanks for all your patience and help in the matter.

Roger

[jasonp]

Originally Posted by roger

Thank you, JasonP

That will take so little time compared to beginning over a Third time.

Thanks for all your patience and help in the matter.

no problem, I know it's frustrating when things don't just work. In general, there is always a way around trouble like this without having to restart from scratch.

jasonp

[roger]

JasonP,

This time instead of corrupted relations, I got the message:
Sat Feb 03 09:08:49 2007 error: read_cycles can't open cycle file
Msieve completed all relations, and gave no warning of errors with those.
I can't find any file with 'cycle' in it, and am lost at what to do.

Thanks,

Roger

[jasonp]

Originally Posted by roger

JasonP,

This time instead of corrupted relations, I got the message:
Sat Feb 03 09:08:49 2007 error: read_cycles can't open cycle file
Msieve completed all relations, and gave no warning of errors with those.
I can't find any file with 'cycle' in it, and am lost at what to do.

This is a message you get when you try to run NFS; what was the command line you used?

jasonp

[roger]

The command line was Msieve -n -nf -s 597~483.dat -v

Thanks

Roger

[jasonp]

Originally Posted by roger

The command line was Msieve -n -nf -s 597~483.dat -v

Now I'm confused; did you restart this factorization from scratch using NFS? If the .dat file contains relations from a QS run, these can't be reused by the number field sieve code (i.e. you can't use -n -nf).

jasonp

[roger]

Sorry for not being clearer.

I have been using the NFS for the factorization since the beginning of this number. It seemed so much faster.

Thanks,

Roger

[jasonp]

Originally Posted by roger

I have been using the NFS for the factorization since the beginning of this number. It seemed so much faster.

What's the time difference?

msieve is looking for the file '<argument of -s>.cyc'; by default it will look for 'msieve.dat.cyc'. Also, -nf expects a filename argument afterwards, or it will pick 'msieve.fb' by default.

If your filename actually has a '~' character in it, this could be confusing windows. When performing an NFS factorization, the temporary files generated during NFS postprocessing get filenames of '<argument of -s>.<something>'

jasonp

[roger]

Thanks, I will change the filename and retry it.

I don't know about the time difference, because I aborted it when on Msieve at 12 minutes (having 10 relations out of a needed 140,000, but I know a lot of that was initializing...). As for NFS : it does somewhere around 11,000 relations an hour (out of 1.862M). My computer is a PIII 550MHz.

Thanks for everything,

Roger

[jasonp]

Originally Posted by roger

I don't know about the time difference, because I aborted it when on Msieve at 12 minutes (having 10 relations out of a needed 140,000, but I know a lot of that was initializing...). As for NFS : it does somewhere around 11,000 relations an hour (out of 1.862M). My computer is a PIII 550MHz.

This can be misleading. For QS, the relation rate starts off very slow and increases to very fast as more and more of the required sieving happens. For NFS, the relation rate starts off high and slows down as more sieving happens. For a 105-digit run the NFS code will probably need 6-7 million relations by the time it's done. Note that if you're using NFS, I highly recommend switching to the latest msieve version, since it contains critical bugfixes.
The NFS postprocessing will also probably requires 150-200MB of RAM.

Are you sure you don't want to let someone finish this run for you? GGNFS on a fast machine can do this job in probably 12 hours.

jasonp