j3dr have no out put

[congcong liu]

Hi all,

I use the command   

mpirun -n 32 --hostfile host-j3dr j3dr    test.lst   test.mrc    --apix 1.08 --sym=c1 --iter=1 --preprocess filter.ctf:type=1:oversample=0 --preprocess xform.centerbyxform --preprocess normalize.mask.circlemean:mask=1:radius=90:masksoft=-8:usescale=1 --postprocess normalize.mask.circlemean:mask=1:radius=90:masksoft=-8 --reconstructor fourier:mode=gauss_2:ctfweight=0 --verbose 2 > run_j3dr.log &

I   j3dr can be found  in the background when I    top      the    therminal.  But there is no  output  in the log file.

Sincerely

Congcong

[Wen Jiang]

Try a smaller number of cpus and see if it is different, for example,

mpirun -n 8

[congcong liu]

Hi Dr. Jiang,

Thank you for your reply!

 When   I use a smaller cpus,    mpirun  -n  8,   I   I can get the log, processing 100,000  particles successfuly.

  But   I have  1,000,000 particles. Using  8  cpus, it will stop when  it insert  100,000particles, and with the err  attached below.

  So I want to use  5 nodes on our cluster  with the command:  

mpirun -n 40 --hostfile host-j3dr j3dr    test.lst   test.mrc    --apix 1.08 --sym=c1 --iter=1 --preprocess filter.ctf:type=1:oversample=0 --preprocess xform.centerbyxform --preprocess normalize.mask.circlemean:mask=1:radius=90:masksoft=-8:usescale=1 --postprocess normalize.mask.circlemean:mask=1:radius=90:masksoft=-8 --reconstructor fourier:mode=gauss_2:ctfweight=0 --verbose 2 > run_j3dr.log &

And the host-j3dr file  is writed like this:

10.1.1.30  slots=8

10.1.1.31  slots=8

10.1.1.32  slots=8

10.1.1.33  slots=8

10.1.1.34  slots=8

When I use the command and this host-j3dr file, programs (j3dr)   can be found  in the background when I    top      the    terminal.  But there is no  output  in the log file.

 Thanks.

Sincerely,

Congcong

在 2019年11月10日星期日 UTC+8下午3:26:30，congcong liu写道：

[Wen Jiang]

The error with 1M particles is due to the underlying EMAN2 IO library which cannot deal with a very large number of particles. A fix is not likely to available soon.

The error with no output is due to some racing issue in MPI communication among different processes. Using a smaller number of CPUs can avoid the issue most of the time.