added F77 parallel test and clean up a little

nc_test4/run_par_test.sh

@@ -2,7 +2,6 @@
 
 # This shell runs some parallel I/O tests.
 
-mpd&
 set -e
 echo
 echo "Testing MPI parallel I/O without netCDF..."

nc_test4/run_valgrind_tests.sh

@@ -25,5 +25,4 @@ for tst in $list; do
 done
 
 echo "SUCCESS!!!"
-
 exit 0

nf_test/Makefile.am

@@ -145,8 +145,9 @@ endif #BUILD_F90
 # Test parallel I/O for F77.
 if TEST_F77
 if TEST_PARALLEL
-check_PROGRAMS += ftst_parallel
+check_PROGRAMS += ftst_parallel ftst_parallel_nasa
 ftst_parallel_SOURCES = ftst_parallel.F
+ftst_parallel_nasa_SOURCES = ftst_parallel_nasa.F
 TESTS += run_f77_par_test.sh
 CLEANFILES += ftst_parallel.nc
 endif # TEST_PARALLEL

nf_test/ftst_parallel_nasa.F

@@ -0,0 +1,142 @@
+!     This is part of the netCDF package. Copyright 2011 University
+!     Corporation for Atmospheric Research/Unidata. See COPYRIGHT file
+!     for conditions of use. 
+
+!     This program tests netCDF-4 parallel I/O from fortran. This
+!     variation of the test was contributed by Dan at NASA. Thanks Dan!
+
+      program ftst_parallel_nasa
+      implicit none
+      include 'netcdf.inc'
+      include 'mpif.h'
+
+      character*(*) FILE_NAME
+      parameter (FILE_NAME = 'ftst_parallel_nasa.nc')
+
+      integer MAX_DIMS
+      parameter (MAX_DIMS = 2)
+      integer NX, NY
+      parameter (NX = 16)
+      parameter (NY = 16)
+      integer NUM_PROC
+      parameter (NUM_PROC = 4)
+      integer ncid, varid, dimids(MAX_DIMS)
+      integer x_dimid, y_dimid
+      real data_out(NY / 2, NX / 2), data_in(NY / 2, NX / 2)
+      integer mode_flag
+      integer x, y, retval
+      integer p, my_rank, ierr
+      integer start(MAX_DIMS), count(MAX_DIMS)
+
+      call MPI_Init(ierr)
+      call MPI_Comm_rank(MPI_COMM_WORLD, my_rank, ierr)
+      call MPI_Comm_size(MPI_COMM_WORLD, p, ierr)
+
+      if (my_rank .eq. 0) then
+         print *, ' '
+         print *, '*** Testing netCDF-4 parallel I/O from F77 again.'
+      endif
+
+!     There must be 4 procs for this test.
+      if (p .ne. 4) then
+         print *, 'This test program must be run on 4 processors.'
+         stop 2
+      endif
+
+!     Create some pretend data.
+      do x = 1, NX / 2
+         do y = 1, NY / 2
+            data_out(y, x) = real(my_rank)
+         end do
+      end do
+
+!     Create the netCDF file.
+      mode_flag = IOR(nf_netcdf4, nf_clobber)
+!     mode_flag = IOR(nf_netcdf4, nf_classic_model)
+      mode_flag = IOR(mode_flag, nf_mpiio)
+      retval = nf_create_par(FILE_NAME, mode_flag, MPI_COMM_WORLD,
+     $     MPI_INFO_NULL, ncid)
+      if (retval .ne. nf_noerr) stop 3
+
+!     Define the dimensions.
+      retval = nf_def_dim(ncid, "x", NX, x_dimid)
+      if (retval .ne. nf_noerr) stop 4
+      retval = nf_def_dim(ncid, "y", NY, y_dimid)
+      if (retval .ne. nf_noerr) stop 5
+      dimids(1) = x_dimid
+      dimids(2) = y_dimid
+
+!     Define the variable.
+      retval = nf_def_var(ncid, "data", NF_FLOAT, MAX_DIMS, dimids,
+     $     varid)
+      if (retval .ne. nf_noerr) stop 6
+
+!     With classic model netCDF-4 file, enddef must be called.
+      retval = nf_enddef(ncid)
+      if (retval .ne. nf_noerr) stop 7
+
+!     Determine what part of the variable will be written for this
+!     processor. It's a checkerboard decomposition.
+      count(1) = NX / 2
+      count(2) = NY / 2
+      if (my_rank .eq. 0) then
+         start(1) = 1
+         start(2) = 1
+      else if (my_rank .eq. 1) then
+         start(1) = NX / 2 + 1
+         start(2) = 1
+      else if (my_rank .eq. 2) then
+         start(1) = 1
+         start(2) = NY / 2 + 1
+      else if (my_rank .eq. 3) then
+         start(1) = NX / 2 + 1
+         start(2) = NY / 2 + 1
+      endif
+
+!     Write this processor's data.
+      retval = nf_put_vara_real(ncid, varid, start, count, data_out)
+      if (retval .ne. nf_noerr) then
+         print*,'Error writing data ', retval
+         print*, NF_STRERROR(retval)
+         stop 8
+      endif
+
+!     Close the file.
+      retval = nf_close(ncid)
+      if (retval .ne. nf_noerr) stop 9
+
+!     Reopen the file.
+      retval = nf_open_par(FILE_NAME, IOR(nf_nowrite, nf_mpiio),
+     $     MPI_COMM_WORLD, MPI_INFO_NULL, ncid)
+      if (retval .ne. nf_noerr) stop 10
+
+!     Set collective access on this variable. This will cause all
+!     reads/writes to happen together on every processor. Fairly
+!     pointless, in this contexct, but I want to at least call this
+!     function once in my testing.
+      retval = nf_var_par_access(ncid, varid, nf_collective)
+      if (retval .ne. nf_noerr) stop 11
+
+!     Read this processor's data.
+      retval = nf_get_vara_real(ncid, varid, start, count, data_in)
+      if (retval .ne. nf_noerr) stop 12
+
+!     Check the data.
+      do x = 1, NX / 2
+         do y = 1, NY / 2
+            if (data_in(y, x) .ne. my_rank) then
+               print*,data_in(y, x), ' NE ', my_rank
+               stop 13
+            endif
+         end do
+      end do
+
+!     Close the file.
+      retval = nf_close(ncid)
+      if (retval .ne. nf_noerr) stop 14
+
+      call MPI_Finalize(ierr)
+
+      if (my_rank .eq. 0) print *,'*** SUCCESS!'
+
+      end program ftst_parallel_nasa

nf_test/run_f77_par_test.sh

@@ -1 +1,12 @@
+#!/bin/sh
+
+# This shell runs some parallel I/O tests for the F77 API.
+
+set -e
+echo "Testing netCDF parallel I/O through the F77 API..."
+
 mpiexec -n 4 ./ftst_parallel
+mpiexec -n 4 ./ftst_parallel_nasa
+
+echo "SUCCESS!!!"
+exit 0

nf_test/run_f90_par_test.sh

@@ -1,7 +1,16 @@
+#!/bin/sh
+
+# This shell runs some parallel I/O tests.
+
 set -e
+echo "Testing netCDF parallel I/O through the F90 API..."
 mpiexec -n 4 ./f90tst_parallel
 mpiexec -n 4 ./f90tst_parallel2
 mpiexec -n 4 ./f90tst_parallel3
 mpiexec -n 8 ./f90tst_nc4_par
 #mpiexec -n 4 ./f90tst_parallel_fill
 
+echo "SUCCESS!!!"
+exit 0
+
+