Msieve LA with openmpi in the Current Age

[EdH]

Originally Posted by VBCurtis

Agree to both on #2; the 18.04 openMPI is likely the problem, and the msieve bug means I shouldn't be trying it on the C207 matrix. My original intent was for you to send an msieve binary, because it didn't occur to me that MPI might be broken.

I discovered the openmpi trouble because I updated some machines and couldn't get them to work in my ecmpi cluster anymore.

[EdH]

Originally Posted by jasonp

If you are working with MPI over multiple machines, can you see if the postprocessing runs to completion using the latest Msieve svn? Greg has reported that it does not (the square root doesn't like the data it's given). It doesn't have to be for a large problem, just a C100 on a 1x2 or 2x2 grid would be enough.

Just in case you haven't looked yet, the trouble seems to appear with revision 1025. 1024 runs fine for me here, while 1025 does not.

[jasonp]

Originally Posted by EdH

Just in case you haven't looked yet, the trouble seems to appear with revision 1025. 1024 runs fine for me here, while 1025 does not.

Yes, 1025 is the merge from the msieve-lacuda branch. The problem is likely not in the linear algebra per se but in the postprocessing that recovers the dependencies. Someday I will have to track it down.

[EdH]

Originally Posted by jasonp

Yes, 1025 is the merge from the msieve-lacuda branch. The problem is likely not in the linear algebra per se but in the postprocessing that recovers the dependencies. Someday I will have to track it down.

At least I know if I want to play any more with mpi I can use 1024 for now. I'm kind of torn between getting some Infiniband gear and trying to get a better handle on CADO-NFS's BWC mpi workings that are supposed to be fine with Gigabit.

If you do want me to run anything else, let me know.