Python Driver for GGNFS and MSIEVE

[WraithX]

Originally Posted by EdH

Edit: I think the trouble might be due to the use of back instead of forward slashes:

ggnfs_cmd = C:\Program Files\Python\python "C:\MathWork\aliqueit\factmsieve.py"

The error you received was because of the space between Program and Files in the above command line. So, you might first want to try putting quotes (") around the python part of your line, like so:

ggnfs_cmd = "C:\Program Files\Python\python" "C:\MathWork\aliqueit\factmsieve.py"

If you still get an error after that, please post back here and we can troubleshoot further.

[smh]

How about putting the path between quotes?

Or use the old 8.3 style, like c:\progra~1\

[EdH]

Thanks Guys,

I guess I was thrown by being too hurried and in thinking that factmsieve.py must have been called fine because of:

error (line 104): CUDA_ERROR_FILE_NOT_FOUND

seeming to indicate Python was that far through the script.

[Mr. Odd]

Originally Posted by EdH

Thanks Guys,

I guess I was thrown by being too hurried and in thinking that factmsieve.py must have been called fine because of:

error (line 104): CUDA_ERROR_FILE_NOT_FOUND

seeming to indicate Python was that far through the script.

Thanks for the help, fixing the "C:\Program Files\" reference worked and GGNFS launches, but I'm still getting that line 104 CUDA error, which is odd as I have another installation of GGNFS that works fine with CUDA turned on. Suggestions?

[jasonp]

msieve with CUDA support looks for a few *.ptx files in the current working directory; the CUDA runtime pops out this error when it doesn't find them. Either the files aren't there or the current directory as specified by python is not what you think it is.

[henryzz]

Are there any files that are used for every factorization? Could I run multiple factorizations from the same folder at once?
edit: silly question there is spairs.out and related files
It might be a nice optional feature to be able to run multiple factorizations in one directory though.

[Jeff Gilchrist]

Originally Posted by henryzz

It might be a nice optional feature to be able to run multiple factorizations in one directory though.

What benefit would that feature give to you? I'm trying to understand why someone would select such an option if it existed.

Jeff.

[henryzz]

Originally Posted by Jeff Gilchrist

What benefit would that feature give to you? I'm trying to understand why someone would select such an option if it existed.

Jeff.

If I ran four factorizations at once in the same directory with one core instead of one with 4 cores then I wouldn't have 3 cores idle during poly selection and filtering. I have however noticed since that with the factmsieve.py script it almost always only runs filtering once unlike factmsieve.pl which ran it many times(sometimes each filtering run taking longer than one sieve block with four cores sieving)

[mdettweiler]

Originally Posted by henryzz

If I ran four factorizations at once in the same directory with one core instead of one with 4 cores then I wouldn't have 3 cores idle during poly selection and filtering. I have however noticed since that with the factmsieve.py script it almost always only runs filtering once unlike factmsieve.pl which ran it many times(sometimes each filtering run taking longer than one sieve block with four cores sieving)

Wouldn't running factmsieve.py 4 times from 4 separate directories do the trick? (You can still use just one copy of the GGNFS binaries and factmsieve.py as long as you kick off each factorization from inside the right directory.)

[henryzz]

Originally Posted by mdettweiler

Wouldn't running factmsieve.py 4 times from 4 separate directories do the trick? (You can still use just one copy of the GGNFS binaries and factmsieve.py as long as you kick off each factorization from inside the right directory.)

Yes need to get around to doing that.

[Jeff Gilchrist]

Originally Posted by henryzz

Yes need to get around to doing that.

That is why I was asking because I figured you could probably do what you needed without having to add support in the script. If you configure factmsieve.py to point to the directory where the msieve/ggnfs binaries are located then you can just call ..\factmsieve.py (or whatever) from your 4 different working directories and be get the same results. That way all your output and data files are nicely separated as well and easier to clean-up.

Request complete, feature already available.