Msieve v1.48 feedback

[jasonp]

Go ahead and PM me the details, I should have time this evening or tomorrow morning.

[xilman]

Originally Posted by jasonp

Go ahead and PM me the details, I should have time this evening or tomorrow morning.

Sent email instead. Hope that's ok.

Paul

[em99010pepe]

Energy went down when I was doing a post-processing ( LA phase) and now I can't resume work.
Here's msieve log.

Thu Mar 10 22:59:46 2011  Msieve v. 1.48
Thu Mar 10 22:59:46 2011  random seeds: e757cb30 67043709
Thu Mar 10 22:59:46 2011  factoring xxxxxxxxxx(234 digits)
Thu Mar 10 22:59:47 2011  searching for 15-digit factors
Thu Mar 10 22:59:48 2011  commencing number field sieve (234-digit input)
Thu Mar 10 22:59:48 2011  elapsed time 00:00:02

Flags used

start /low /min msieve.exe -i integer.ini -s integer.dat -nf integer.fb -nc2 -nc3 -ncr -v -t 4

Files available:
*.dat
*.bak.chk
*.chk
*.cyc
*.mat

[jasonp]

I take it you don't have console output from the run? Looks like the error is only printed to the screen.

Unless the matrix was actually rebuilt by accident, the current code saves two checkpoint files, so you may want to try restarting from the older one if it exists.

you can rename the .chk file and then rename the .chk.bak file to .chk to try this out.

[em99010pepe]

Originally Posted by jasonp

I take it you don't have console output from the run? Looks like the error is only printed to the screen.

I run under windows...

Originally Posted by jasonp

Unless the matrix was actually rebuilt by accident, the current code saves two checkpoint files, so you may want to try restarting from the older one if it exists.

Made that, same behave.

Originally Posted by jasonp

you can rename the .chk file and then rename the .chk.bak file to .chk to try this out.

Done. Same behave.

[jasonp]

Could you run the command line in a windows command prompt, without the 'start /low /min' so we can see the console output?

[em99010pepe]

Originally Posted by jasonp

Could you run the command line in a windows command prompt, without the 'start /low /min' so we can see the console output?

Sorry for the delay in the answer, I went to sleep last night. Here you go:

Msieve v. 1.48
Fri Mar 11 07:23:05 2011
random seeds: 89702e40 3bd083cc
factoring 6781744011973536532556183189715932489327872220859163365620930108658414
75124480105672401904335969730470024772368102507025579331092102993264883430426192
39774183409536873843517888777485153682357422682030299183493881998271868596989822
4659 (234 digits)
searching for 15-digit factors
commencing number field sieve (234-digit input)
R0: -11283389295712565329310524347604879658147
R1:  1
A0: -1
A1:  0
A2:  0
A3:  0
A4:  0
A5:  0
A6:  227
skew 0.40, size 3.972e-012, alpha 2.243, combined = 3.296e-013 rroots = 2

commencing linear algebra
error: cannot open matrix file

Now what, I need to re-start the all process? Is there away to recover the matrix file?

[jasonp]

This doesn't sound too bad. Can you do a dir on the current directory? The next thing to check is how large the files are. Is this an RSALS postprocessing job? Because if you don't have the actual relations then you shouldn't be running with -nc2, since that will try to build the matrix (which you already have).

[em99010pepe]

Originally Posted by jasonp

This doesn't sound too bad. Can you do a dir on the current directory? The next thing to check is how large the files are. Is this an RSALS postprocessing job? Because if you don't have the actual relations then you shouldn't be running with -nc2, since that will try to build the matrix (which you already have).

I found the problem. The batch file flags had one command wrong, name of the fb file misspelled. Sorry for the trouble. Yesterday night I was with my eyes tired so I couldn't see right ( I use glasses). It's running now..

Carlos

[jasonp]

Glad to help :)

[xilman]

No, not the movie.

Would it be a good idea if the square root phase reported the reason why each dependency gave the result it did? At the moment, unless all the factors are available it seems to move on silently to the next .

The CWI code written by Peter Montgomery reports any reason why the square root phase may require further iterations. IIRC, it may report that the GCD is 1, or is N (the number being factored) or that only one factor is prime. In principle, I guess that both factors could be composite and further iterations are need to split them both but I've never seen it happen.

I happen rather to like the extra information, especially to see intermediate factors asap for a > 2-way split. Should this behaviour also be implemented in msieve?


Paul