variables

[sbshewale@gmail_com]

I am running cam5 (CESM1_0) in parrallel mode (for 1 day) and submitting job directly on commandline asmpirun -np 24 ./cam >& run.out &my question iseven if I have given 24 processors to run the job, but How I will come to know, that how many processors it is taking to run job?also another doubtHow many no. of variable  it should create  by default (in .nc file)Actually I am having subfile, but somehow it may be having some problemso, If anyone is having subfile to submit the job (by qsub/bsub),I will be happy to use it.Thanks

 

[fischer]

You can see the processor layout in the log file that was created by your run.  It will look something like. (seq_comm_setcomm)  initialize ID ( 17 ATM             ) pelist   =     0    31     1 ( npes =    32) ( nthreads =  2)

 You can also find the number of variables being created by default in the history fileby looking in the log file.  The section of the log file should look something like  FLDLST: History file            1  contains          255  fields
  Write frequency:                            3
  Filename specifier:              %c.cam.h%t.%y-%m-%d-%s.nc
  Output precision:                double
  Number of time samples per file:            1
  Included fields are:

Chris

 

[sbshewale@gmail_com]

Thanks for replyI am have some questions,(seq_comm_setcomm)  initialize ID ( 17 ATM             ) pelist   =     0    31     1 ( npes =    32) ( nthreads =  2)so does it mean, it is using 32 no. of procassor to run the job ? (what is then 17 ATM ?)If I mention thread also at the time of configuration,  will it make any difference?I mean run will complete in less time?mentioning both ntasks and nthread, run will be in hybrid mode, thenshould I use the same command (mpirun -np 24 ./cam >& run.out &) to execute the job?also, if I mention both ntasks and nthreads as 48 (-ntasks 48 -nthreads 48) at the time of configuration.how many no of processors should I mention at the time of job run? or should I give same 48 no pf procassors (mpirun -np 48 ./cam)?Actually I want to submit the job through script (subfile)If someone knows, will you Please guide the sameThanks

 

[eaton]

The line from the output log says that the atm is using 32 tasks with 2 threads per task.  17 is an internal ID used by MPI.  Threading will not necessarily increase performance.  That is very machine specific.  Usually at low task count a pure MPI (no threads) configuration is the best option.