T3E MPI-C & MPI-F90 Source, Ouput, and Batch Job Script Examples:

• 2D Laplace Equation by Jacobi Method Interactive Code Example:
  1. T3E MPI Laplace Jacobi Computing Code (MPT Revised):

     laplace.mpi_c.c
     OR
     laplace.mpi_f.f

  2. T3E MPI Laplace Jacobi Computing Output:

     laplace.mpi_c.output for Nonconverging 1000 interations in C
     OR
     laplace.mpi_c3000.output for Converging 3000 interations in C
     OR
     laplace.mpi_f.output for Nonconverging 1000 interations in F90
     OR
     laplace.mpi_f.output for Converging 4000 interations in F90

  3. T3E MPI NQS Job Script used with Source in cpgm.c or fpgm.f with Data in cdata or fdata, respectively:

     cpgm.job for C
     OR
     fpgm.job for F90

  4. Note: The C and F90 programs use different problem examples, the F90 one using an problem that has a exact second degree polynomial solution.

• Trapezoidal Rule Code Example (Adapted from Pacheco's PPMPI Code):
  1. T3E MPI Trapezoidal Rule Computing Code:  

     trap_mpi_c.c for C
     OR
     trap_mpi_f.f for F90

  2. T3E MPI C Trapezoidal Rule Computing Output:  

     trap_mpi_c.output for C
     OR
     trap_mpi_f.output for F90

  3. Note: Compile Step for code cpgm.c OR fpgm.f:

     
     cc -O3 -h report=isvf -Xm -o cpgm cpgm.c &' for C
     OR
     `f90 -O3 -r3 -Xm -o fpgm fpgm.f &' for F90

  4. Execute Step:

     
     `mpprun -n4 cpgm>cpgm.output&' for C
     OR
     `mpprun -n4 fpgm>fpgm.output&' for F90
     but will run interactively with up to 7 processors.

• PI Code Example:
  1. Click for T3D MPI Pi Computing Code:  

     pi_mpi.c for C
     OR
     pi_mpi.f for F90

  2. Click for T3D MPI Pi Computing Output:  

     pi_mpi.c.out for C
     OR
     pi_mpi.f.out for F90

  3. FOR Data:

     pi_mpi.data for C or F90

• SDSC/NPACI or PSC Cray T3E MPI Code NQS Generic Batch Job Script (Use for Big Computation or Big Memory Jobs And for 8 to 128 Processors):
  1. Get MCS572 T3E Generic Script File for 4 Processors Compiles & Runs:

     cpgm.job; for C
     OR
     fpgm.job; for F90

     See also some example using more than 4 processors (the 4 processor example is meant to be a test case):

     cpgm8.job;
     OR
     fpgm16.job;
     OR
     cpgm32.job;

  2. Copy Your MPI Source as Generic File `cpgm.c' OR `fpgm.f' (Name Used in Job Script);

  3. Copy (or Create by vi) MPI Input Data as Generic File `cdata' OR `fdata' (Name Used in Job Script; If Code Does Not Require Redirected Data Input, Then Any File Called `cdata' OR `fdata' Will Do);

  4. Enter `qsub cpgm.job' OR `qsub fpgm.job' Queue Submit Command on the T3E Command Line (Submit Job to Batch Queue and Executes Using Temporary Directory Just for This Job, But Output Should Appear in Submitted Directory);

  5. Enter `qstat -a' Queue Status Commmand Until "no batch queue entries" Found which means your job is in the RUN Queue ; ("no pipe queue entries" means you job is not yet in the RUN Queue in the queued QUE Queue); OR use `qstat -b' for a more total view of batch queues;

  6. Enter `vi cpgm[n].output' OR `vi fpgm[n].output', where `[n]' is the selected number of processors in the Job Script, Renaming Output If Desired and/or FTP Back to UIC for Printing.

• T3D MPI Fortran Laplace Red-Black-SOR Computing Code:
  1. Click for T3D MPI Fortran Laplace Red-Black-SOR Computing Code:   ***NOT YET READY*** laplace_sorrbo_mpi_ornl.f

  2. Click for T3D MPI Fortran Laplace Red-Black-SOR Computing Output:   laplace_sorrbo_mpi_ornl.f.out

T90 MPI Example using MPIRUN Execution:

1. Tested on NPACI-SDSC T90 using the Pi calculation with modified source codes

   pi_mpi_t90.c
   AND
   pi_mpi_t90.f,

   which have been modified by changing the "real*8" datatype to "double precision" to make the datatypes between F90 and MPI consistent to avoid wordsize errors (F90 datatypes differ between T90 and T3D), and the T3E "fprintf" synchronization error has been fixed.

2. Instructions assumed that source files are copied into dummy or generic names, e.g.,

   "cp pi_mpi_t90.c cpgm.c"
   OR
   "cp pi_mpi_t90.f fpgm.f".

3. Compile by

   "cc -O3 -h scalar2 -h report=isvf -o cpgm cpgm.c &"
   OR
   "f90 -O3 -r3 -o fpgm fpgm.f &",

   noting that the malleable "-Xm" or "-X[n]" processor number compiler option is not supported on the T90.

4. Execute on 4 processors by

   "mpirun -np 4 cpgm < cdata >& cpgm.output&"
   OR
   "mprun -np 4 fpgm < fdata >& fpgm.output&",

   noting that the processor number option is "-np [n]" using the "mpirun" command which works on both T90 and T3E, but is different from the "mpprun" command which only works on the T3E; for executing on a different number of processors, just change "[n]" to the actual number.

5. For the data files used for the PI example codes were

   cdata
   OR
   fdata.

6. For repeated compilations-executions, you need to remove output files to avoid protection overwrite errors:

   "rm cpgm cpgm.V cpgm.output"
   OR
   "rm fpgm fpgm.lst fpgm.output"

   or just repeated executions, you need just to remove the old output file:

   "rm cpgm.output"
   OR
   "rm fpgm.output"