mirror of
https://github.com/Unidata/netcdf-c.git
synced 2024-11-21 03:13:42 +08:00
826 lines
22 KiB
C
826 lines
22 KiB
C
/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
|
|
* Copyright by the Board of Trustees of the University of Illinois. *
|
|
* All rights reserved. *
|
|
* *
|
|
* This file is part of HDF5. The full HDF5 copyright notice, including *
|
|
* terms governing use, modification, and redistribution, is contained in *
|
|
* the files COPYING and Copyright.html. COPYING can be found at the root *
|
|
* of the source code distribution tree; Copyright.html can be found at the *
|
|
* root level of an installed copy of the electronic HDF5 document set and *
|
|
* is linked from the top-level documents page. It can also be found at *
|
|
* http://hdf.ncsa.uiuc.edu/HDF5/doc/Copyright.html. If you do not have *
|
|
* access to either file, you may request a copy from hdfhelp@ncsa.uiuc.edu. *
|
|
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
|
|
|
|
|
|
/*
|
|
* Main driver of the Parallel NetCDF4 tests
|
|
*
|
|
*/
|
|
|
|
#include <nc_tests.h>
|
|
|
|
#define FILE_NAME "tst_parallel3.nc"
|
|
|
|
/*2,3,4 dimensional test, the first dimension is unlimited, time.
|
|
*/
|
|
|
|
#define NDIMS1 2
|
|
#define NDIMS2 4
|
|
#define DIMSIZE /*4 */ 768*2
|
|
#define DIMSIZE2 4
|
|
#define DIMSIZE3 4
|
|
#define TIMELEN 1
|
|
|
|
/*BIGFILE, >2G, >4G, >8G file
|
|
big file is created but no actually data is written
|
|
Dimensional size is defined inside the function
|
|
*/
|
|
|
|
#define ATTRNAME1 "valid_range"
|
|
#define ATTRNAME2 "scale_factor"
|
|
#define ATTRNAME3 "title"
|
|
|
|
/* The number of processors should be a good number for the
|
|
dimension to be divided evenly, the best set of number of processor
|
|
should be 2 power n. However, for NetCDF4 tests, the following numbers
|
|
are generally treated as good numbers:
|
|
1,2,3,4,6,8,12,16,24,32,48,64,96,128,192,256
|
|
|
|
The maximum number of processor is 256.*/
|
|
|
|
int test_pio(int);
|
|
int test_pio_attr(int);
|
|
int test_pio_big(int);
|
|
int test_pio_hyper(int);
|
|
int test_pio_extend(int);
|
|
|
|
char* getenv_all(MPI_Comm comm, int root, const char* name);
|
|
int facc_type;
|
|
int facc_type_open;
|
|
char file_name[NC_MAX_NAME + 1];
|
|
|
|
int main(int argc, char **argv)
|
|
{
|
|
int mpi_size, mpi_rank; /* mpi variables */
|
|
int i;
|
|
int NUMP[16] ={1,2,3,4,6,8,12,16,24,32,48,64,96,128,192,256};
|
|
int size_flag = 0;
|
|
|
|
/* Un-buffer the stdout and stderr */
|
|
setbuf(stderr, NULL);
|
|
setbuf(stdout, NULL);
|
|
|
|
MPI_Init(&argc, &argv);
|
|
MPI_Comm_size(MPI_COMM_WORLD, &mpi_size);
|
|
MPI_Comm_rank(MPI_COMM_WORLD, &mpi_rank);
|
|
|
|
if (mpi_rank == 0)
|
|
printf("\n*** Testing more advanced parallel access.\n");
|
|
|
|
for (i = 0; i < 16; i++){
|
|
if(mpi_size == NUMP[i])
|
|
{
|
|
size_flag = 1;
|
|
break;
|
|
}
|
|
}
|
|
if(!size_flag){
|
|
printf("mpi_size is wrong\n");
|
|
printf(" The number of processor must be chosen from\n");
|
|
printf(" 1,2,3,4,6,8,12,16,24,32,48,64,96,128,192,256 \n");
|
|
return -1;
|
|
}
|
|
|
|
facc_type = NC_NETCDF4|NC_MPIIO;
|
|
facc_type_open = NC_MPIIO;
|
|
|
|
/* Create file name. */
|
|
sprintf(file_name, "%s/%s", TEMP_LARGE, FILE_NAME);
|
|
|
|
/* Test NetCDF4 with MPI-IO driver */
|
|
if (mpi_rank == 0)
|
|
printf("*** Testing parallel IO for raw-data with MPI-IO (driver)...");
|
|
if(test_pio(NC_INDEPENDENT)!=0) ERR;
|
|
if(test_pio(NC_COLLECTIVE)!=0) ERR;
|
|
if (mpi_rank == 0)
|
|
SUMMARIZE_ERR;
|
|
|
|
if (mpi_rank == 0)
|
|
printf("*** Testing parallel IO for meta-data with MPI-IO (driver)...");
|
|
if(test_pio_attr(NC_INDEPENDENT)!=0) ERR;
|
|
if(test_pio_attr(NC_COLLECTIVE)!=0) ERR;
|
|
if (mpi_rank == 0)
|
|
SUMMARIZE_ERR;
|
|
|
|
if (mpi_rank == 0)
|
|
printf("*** Testing parallel IO for different hyperslab selections with MPI-IO (driver)...");
|
|
if(test_pio_hyper(NC_INDEPENDENT)!=0)ERR;
|
|
if(test_pio_hyper(NC_COLLECTIVE)!=0) ERR;
|
|
if (mpi_rank == 0)
|
|
SUMMARIZE_ERR;
|
|
|
|
if (mpi_rank == 0)
|
|
printf("*** Testing parallel IO for extending variables with MPI-IO (driver)...");
|
|
if(test_pio_extend(NC_COLLECTIVE)!=0) ERR;
|
|
if (mpi_rank == 0)
|
|
SUMMARIZE_ERR;
|
|
|
|
/* Note: When the MPI-POSIX VFD is not compiled in to HDF5, the NC_MPIPOSIX
|
|
* flag will be aliased to the NC_MPIIO flag within the library, and
|
|
* therefore this test will exercise the aliasing, with the MPI-IO VFD,
|
|
* under that configuration. -QAK
|
|
*/
|
|
if (mpi_rank == 0)
|
|
printf("*** Testing parallel IO for raw-data with MPIPOSIX-IO (driver)...");
|
|
facc_type = NC_NETCDF4|NC_MPIPOSIX;
|
|
facc_type_open = NC_MPIPOSIX;
|
|
if(test_pio(NC_INDEPENDENT)!=0) ERR;
|
|
if(test_pio(NC_COLLECTIVE)!=0) ERR;
|
|
if (mpi_rank == 0)
|
|
SUMMARIZE_ERR;
|
|
|
|
if (mpi_rank == 0)
|
|
printf("*** Testing parallel IO for meta-data with MPIPOSIX-IO (driver)...");
|
|
if(test_pio_attr(NC_INDEPENDENT)!=0) ERR;
|
|
if(test_pio_attr(NC_COLLECTIVE)!=0) ERR;
|
|
if (mpi_rank == 0)
|
|
SUMMARIZE_ERR;
|
|
|
|
if (mpi_rank == 0)
|
|
printf("*** Testing parallel IO for different hyperslab selections "
|
|
"with MPIPOSIX-IO (driver)...");
|
|
if(test_pio_hyper(NC_INDEPENDENT)!=0)ERR;
|
|
if(test_pio_hyper(NC_COLLECTIVE)!=0) ERR;
|
|
if (mpi_rank == 0)
|
|
SUMMARIZE_ERR;
|
|
|
|
if (mpi_rank == 0)
|
|
printf("*** Testing parallel IO for extending variables with MPIPOSIX-IO (driver)...");
|
|
if(test_pio_extend(NC_COLLECTIVE)!=0) ERR;
|
|
if (mpi_rank == 0)
|
|
SUMMARIZE_ERR;
|
|
|
|
/* if(!getenv_all(MPI_COMM_WORLD,0,"NETCDF4_NOCLEANUP")) */
|
|
remove(file_name);
|
|
MPI_Finalize();
|
|
|
|
if (mpi_rank == 0)
|
|
FINAL_RESULTS;
|
|
return 0;
|
|
}
|
|
|
|
/* Both read and write will be tested */
|
|
int test_pio(int flag)
|
|
{
|
|
/* MPI stuff. */
|
|
int mpi_size, mpi_rank;
|
|
MPI_Comm comm = MPI_COMM_WORLD;
|
|
MPI_Info info = MPI_INFO_NULL;
|
|
|
|
/* Netcdf-4 stuff. */
|
|
int ncid;
|
|
int nvid,uvid;
|
|
int rvid;
|
|
unsigned m,k,j,i;
|
|
|
|
/* two dimensional integer data test */
|
|
int dimids[NDIMS1];
|
|
size_t start[NDIMS1];
|
|
size_t count[NDIMS1];
|
|
|
|
int *data;
|
|
int *tempdata;
|
|
int *rdata;
|
|
int *temprdata;
|
|
|
|
/* four dimensional integer data test,
|
|
time dimension is unlimited.*/
|
|
int dimuids[NDIMS2];
|
|
size_t ustart[NDIMS2];
|
|
size_t ucount[NDIMS2];
|
|
|
|
int *udata;
|
|
int *tempudata;
|
|
int *rudata;
|
|
int *temprudata;
|
|
|
|
/* Initialize MPI. */
|
|
MPI_Comm_size(MPI_COMM_WORLD, &mpi_size);
|
|
MPI_Comm_rank(MPI_COMM_WORLD, &mpi_rank);
|
|
|
|
/* Create a parallel netcdf-4 file. */
|
|
if (nc_create_par(file_name, facc_type, comm, info, &ncid)) ERR;
|
|
|
|
/* The first case is two dimensional variables, no unlimited dimension */
|
|
|
|
/* Create two dimensions. */
|
|
if (nc_def_dim(ncid, "d1", DIMSIZE2, dimids)) ERR;
|
|
if (nc_def_dim(ncid, "d2", DIMSIZE, &dimids[1])) ERR;
|
|
|
|
/* Create one var. */
|
|
if (nc_def_var(ncid, "v1", NC_INT, NDIMS1, dimids, &nvid)) ERR;
|
|
|
|
if (nc_enddef(ncid)) ERR;
|
|
|
|
/* Set up slab for this process. */
|
|
start[0] = 0;
|
|
start[1] = mpi_rank * DIMSIZE/mpi_size;
|
|
count[0] = DIMSIZE2;
|
|
count[1] = DIMSIZE/mpi_size;
|
|
|
|
/* start parallel netcdf4 */
|
|
if (nc_var_par_access(ncid, nvid, flag)) ERR;
|
|
|
|
if (!(data = malloc(sizeof(int)*count[1]*count[0]))) ERR;
|
|
tempdata = data;
|
|
for (j = 0; j < count[0]; j++){
|
|
for (i = 0; i < count[1]; i++)
|
|
{
|
|
*tempdata = mpi_rank * (j + 1);
|
|
tempdata++;
|
|
}
|
|
}
|
|
|
|
/* Write two dimensional integer data */
|
|
if (nc_put_vara_int(ncid, nvid, start, count, data)) ERR;
|
|
free(data);
|
|
|
|
/* Case 2: create four dimensional integer data,
|
|
one dimension is unlimited. */
|
|
|
|
/* Create four dimensions. */
|
|
if (nc_def_dim(ncid, "ud1", NC_UNLIMITED, dimuids)) ERR;
|
|
if (nc_def_dim(ncid, "ud2", DIMSIZE3, &dimuids[1])) ERR;
|
|
if (nc_def_dim(ncid, "ud3", DIMSIZE2, &dimuids[2])) ERR;
|
|
if (nc_def_dim(ncid, "ud4", DIMSIZE, &dimuids[3])) ERR;
|
|
|
|
/* Create one var. */
|
|
if (nc_def_var(ncid, "uv1", NC_INT, NDIMS2, dimuids, &uvid)) ERR;
|
|
|
|
if (nc_enddef(ncid)) ERR;
|
|
|
|
/* Set up selection parameters */
|
|
ustart[0] = 0;
|
|
ustart[1] = 0;
|
|
ustart[2] = 0;
|
|
ustart[3] = DIMSIZE*mpi_rank/mpi_size;
|
|
ucount[0] = TIMELEN;
|
|
ucount[1] = DIMSIZE3;
|
|
ucount[2] = DIMSIZE2;
|
|
ucount[3] = DIMSIZE/mpi_size;
|
|
|
|
/* Access parallel */
|
|
if (nc_var_par_access(ncid, uvid, flag)) ERR;
|
|
|
|
/* Create phony data. */
|
|
if (!(udata = malloc(ucount[0]*ucount[1]*ucount[2]*ucount[3]*sizeof(int)))) ERR;
|
|
tempudata = udata;
|
|
for( m=0; m<ucount[0];m++)
|
|
for( k=0; k<ucount[1];k++)
|
|
for (j=0; j<ucount[2];j++)
|
|
for (i=0; i<ucount[3]; i++)
|
|
{
|
|
*tempudata = (1+mpi_rank)*2*(j+1)*(k+1)*(m+1);
|
|
tempudata++;
|
|
}
|
|
|
|
/* Write slabs of phoney data. */
|
|
if (NC_INDEPENDENT == flag) {
|
|
int res;
|
|
res = nc_put_vara_int(ncid, uvid, ustart, ucount, udata);
|
|
if(res != NC_ECANTEXTEND) ERR;
|
|
}
|
|
else {
|
|
if (nc_put_vara_int(ncid, uvid, ustart, ucount, udata)) ERR;
|
|
}
|
|
free(udata);
|
|
|
|
/* Close the netcdf file. */
|
|
if (nc_close(ncid)) ERR;
|
|
|
|
if (nc_open_par(file_name, facc_type_open, comm, info, &ncid)) ERR;
|
|
|
|
/* Case 1: read two-dimensional variables, no unlimited dimension */
|
|
/* Set up slab for this process. */
|
|
start[0] = 0;
|
|
start[1] = mpi_rank * DIMSIZE/mpi_size;
|
|
count[0] = DIMSIZE2;
|
|
count[1] = DIMSIZE/mpi_size;
|
|
|
|
if (nc_inq_varid(ncid, "v1", &rvid)) ERR;
|
|
|
|
if (nc_var_par_access(ncid, rvid, flag)) ERR;
|
|
|
|
if (!(rdata = malloc(sizeof(int)*count[1]*count[0]))) ERR;
|
|
if (nc_get_vara_int(ncid, rvid, start, count, rdata)) ERR;
|
|
|
|
temprdata = rdata;
|
|
for (j=0; j<count[0];j++){
|
|
for (i=0; i<count[1]; i++){
|
|
if(*temprdata != mpi_rank*(j+1))
|
|
{
|
|
ERR_RET;
|
|
break;
|
|
}
|
|
temprdata++;
|
|
}
|
|
}
|
|
free(rdata);
|
|
|
|
/* Case 2: read four dimensional data, one dimension is unlimited. */
|
|
|
|
/* set up selection parameters */
|
|
ustart[0] = 0;
|
|
ustart[1] = 0;
|
|
ustart[2] = 0;
|
|
ustart[3] = DIMSIZE*mpi_rank/mpi_size;
|
|
ucount[0] = TIMELEN;
|
|
ucount[1] = DIMSIZE3;
|
|
ucount[2] = DIMSIZE2;
|
|
ucount[3] = DIMSIZE/mpi_size;
|
|
|
|
/* Inquiry the data */
|
|
/* (NOTE: This variable isn't written out, when access is independent) */
|
|
if (NC_INDEPENDENT != flag) {
|
|
if (nc_inq_varid(ncid, "uv1", &rvid)) ERR;
|
|
|
|
/* Access the parallel */
|
|
if (nc_var_par_access(ncid, rvid, flag)) ERR;
|
|
|
|
if (!(rudata = malloc(ucount[0]*ucount[1]*ucount[2]*ucount[3]*sizeof(int)))) ERR;
|
|
temprudata = rudata;
|
|
|
|
/* Read data */
|
|
if (nc_get_vara_int(ncid, rvid, ustart, ucount, rudata)) ERR;
|
|
|
|
for(m = 0; m < ucount[0]; m++)
|
|
for(k = 0; k < ucount[1]; k++)
|
|
for(j = 0; j < ucount[2]; j++)
|
|
for(i = 0; i < ucount[3]; i++)
|
|
{
|
|
if(*temprudata != (1+mpi_rank)*2*(j+1)*(k+1)*(m+1))
|
|
ERR_RET;
|
|
temprudata++;
|
|
}
|
|
|
|
free(rudata);
|
|
}
|
|
|
|
/* Close the netcdf file. */
|
|
if (nc_close(ncid)) ERR;
|
|
|
|
return 0;
|
|
}
|
|
|
|
|
|
/* Attributes: both read and write will be tested for parallel NetCDF*/
|
|
|
|
int test_pio_attr(int flag)
|
|
{
|
|
/* MPI stuff. */
|
|
int mpi_size, mpi_rank;
|
|
MPI_Comm comm = MPI_COMM_WORLD;
|
|
MPI_Info info = MPI_INFO_NULL;
|
|
|
|
/* Netcdf-4 stuff. */
|
|
int ncid;
|
|
int nvid;
|
|
int j, i;
|
|
|
|
double rh_range[2];
|
|
static char title[] = "parallel attr to netCDF";
|
|
nc_type st_type,vr_type;
|
|
size_t vr_len,st_len;
|
|
size_t orivr_len;
|
|
double *vr_val;
|
|
char *st_val;
|
|
|
|
/* two dimensional integer data*/
|
|
int dimids[NDIMS1];
|
|
size_t start[NDIMS1];
|
|
size_t count[NDIMS1];
|
|
int *data;
|
|
int *tempdata;
|
|
|
|
/* Initialize MPI. */
|
|
MPI_Comm_size(MPI_COMM_WORLD, &mpi_size);
|
|
MPI_Comm_rank(MPI_COMM_WORLD, &mpi_rank);
|
|
|
|
/* Create a parallel netcdf-4 file. */
|
|
/* nc_set_log_level(NC_TURN_OFF_LOGGING); */
|
|
/* nc_set_log_level(3);*/
|
|
|
|
if (nc_create_par(file_name, facc_type, comm, info, &ncid)) ERR;
|
|
|
|
/* Create a 2-D variable so that an attribute can be added. */
|
|
if (nc_def_dim(ncid, "d1", DIMSIZE2, dimids)) ERR;
|
|
if (nc_def_dim(ncid, "d2", DIMSIZE, &dimids[1])) ERR;
|
|
|
|
/* Create one var. */
|
|
if (nc_def_var(ncid, "v1", NC_INT, NDIMS1, dimids, &nvid)) ERR;
|
|
|
|
orivr_len = 2;
|
|
rh_range[0] = 1.0;
|
|
rh_range[1] = 1000.0;
|
|
|
|
/* Write attributes of a variable */
|
|
|
|
if (nc_put_att_double (ncid, nvid, "valid_range", NC_DOUBLE,
|
|
orivr_len, rh_range)) ERR;
|
|
|
|
if (nc_put_att_text (ncid, nvid, "title", strlen(title),
|
|
title)) ERR;
|
|
|
|
/* Write global attributes */
|
|
if (nc_put_att_double (ncid, NC_GLOBAL, "g_valid_range", NC_DOUBLE,
|
|
orivr_len, rh_range)) ERR;
|
|
if (nc_put_att_text (ncid, NC_GLOBAL, "g_title", strlen(title), title)) ERR;
|
|
|
|
if (nc_enddef(ncid)) ERR;
|
|
|
|
/* Set up slab for this process. */
|
|
start[0] = 0;
|
|
start[1] = mpi_rank * DIMSIZE/mpi_size;
|
|
count[0] = DIMSIZE2;
|
|
count[1] = DIMSIZE/mpi_size;
|
|
|
|
/* Access parallel */
|
|
if (nc_var_par_access(ncid, nvid, flag)) ERR;
|
|
|
|
/* Allocating data */
|
|
data = malloc(sizeof(int)*count[1]*count[0]);
|
|
tempdata = data;
|
|
for(j = 0; j < count[0]; j++)
|
|
for (i = 0; i < count[1]; i++)
|
|
{
|
|
*tempdata = mpi_rank * (j + 1);
|
|
tempdata++;
|
|
}
|
|
|
|
if (nc_put_vara_int(ncid, nvid, start, count, data)) ERR;
|
|
free(data);
|
|
|
|
/* Close the netcdf file. */
|
|
if (nc_close(ncid)) ERR;
|
|
|
|
/* Read attributes */
|
|
if (nc_open_par(file_name, facc_type_open, comm, info, &ncid)) ERR;
|
|
|
|
/* Set up slab for this process. */
|
|
start[0] = 0;
|
|
start[1] = mpi_rank * DIMSIZE/mpi_size;
|
|
count[0] = DIMSIZE2;
|
|
count[1] = DIMSIZE/mpi_size;
|
|
|
|
/* Inquiry variable */
|
|
if (nc_inq_varid(ncid, "v1", &nvid)) ERR;
|
|
|
|
/* Access parallel */
|
|
if (nc_var_par_access(ncid, nvid, flag)) ERR;
|
|
|
|
/* Inquiry attribute */
|
|
if (nc_inq_att (ncid, nvid, "valid_range", &vr_type, &vr_len)) ERR;
|
|
|
|
/* check stuff */
|
|
if(vr_type != NC_DOUBLE || vr_len != orivr_len) ERR;
|
|
|
|
vr_val = (double *) malloc(vr_len * sizeof(double));
|
|
|
|
/* Get variable attribute values */
|
|
if (nc_get_att_double(ncid, nvid, "valid_range", vr_val)) ERR;
|
|
|
|
/* Check variable attribute value */
|
|
for(i = 0; i < vr_len; i++)
|
|
if (vr_val[i] != rh_range[i])
|
|
ERR_RET;
|
|
free(vr_val);
|
|
|
|
/* Inquiry global attribute */
|
|
if (nc_inq_att (ncid, NC_GLOBAL, "g_valid_range", &vr_type, &vr_len)) ERR;
|
|
|
|
/* Check stuff. */
|
|
if(vr_type != NC_DOUBLE || vr_len != orivr_len) ERR;
|
|
|
|
/* Obtain global attribute value */
|
|
vr_val = (double *) malloc(vr_len * sizeof(double));
|
|
if (nc_get_att_double(ncid, NC_GLOBAL, "g_valid_range", vr_val)) ERR;
|
|
|
|
/* Check global attribute value */
|
|
for(i = 0; i < vr_len; i++)
|
|
if (vr_val[i] != rh_range[i]) ERR_RET;
|
|
free(vr_val);
|
|
|
|
/* Inquiry string attribute of a variable */
|
|
if (nc_inq_att (ncid, nvid, "title", &st_type, &st_len)) ERR;
|
|
|
|
/* check string attribute length */
|
|
if(st_len != strlen(title)) ERR_RET;
|
|
|
|
/* Check string attribute type */
|
|
if(st_type != NC_CHAR) ERR_RET;
|
|
|
|
/* Allocate meory for string attribute */
|
|
st_val = (char *) malloc(st_len * (sizeof(char)));
|
|
|
|
/* Obtain variable string attribute value */
|
|
if (nc_get_att_text(ncid, nvid,"title", st_val)) ERR;
|
|
|
|
/*check string value */
|
|
if(strncmp(st_val,title,st_len)) {
|
|
free(st_val);
|
|
ERR_RET;
|
|
}
|
|
free(st_val);
|
|
|
|
/*Inquiry global attribute */
|
|
if (nc_inq_att (ncid, NC_GLOBAL, "g_title", &st_type, &st_len)) ERR;
|
|
|
|
/* check attribute length*/
|
|
if(st_len != strlen(title)) ERR_RET;
|
|
|
|
/*check attribute type*/
|
|
if(st_type != NC_CHAR) ERR_RET;
|
|
|
|
/* obtain global string attribute value */
|
|
st_val = (char*)malloc(st_len*sizeof(char));
|
|
if (nc_get_att_text(ncid, NC_GLOBAL,"g_title", st_val)) ERR;
|
|
|
|
/* check attribute value */
|
|
if(strncmp(st_val,title,st_len)){
|
|
free(st_val);
|
|
ERR_RET;
|
|
}
|
|
free(st_val);
|
|
|
|
/* Close the netcdf file. */
|
|
if (nc_close(ncid)) ERR;
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* test different hyperslab settings */
|
|
int test_pio_hyper(int flag){
|
|
|
|
/* MPI stuff. */
|
|
int mpi_size, mpi_rank;
|
|
int res = NC_NOERR;
|
|
MPI_Comm comm = MPI_COMM_WORLD;
|
|
MPI_Info info = MPI_INFO_NULL;
|
|
|
|
/* Netcdf-4 stuff. */
|
|
int ncid;
|
|
int nvid;
|
|
int rvid;
|
|
int j, i;
|
|
|
|
/* two dimensional integer data test */
|
|
int dimids[NDIMS1];
|
|
size_t start[NDIMS1], count[NDIMS1];
|
|
int *data;
|
|
int *tempdata;
|
|
int *rdata;
|
|
int *temprdata;
|
|
int count_atom;
|
|
|
|
|
|
/* Initialize MPI. */
|
|
MPI_Comm_size(MPI_COMM_WORLD, &mpi_size);
|
|
MPI_Comm_rank(MPI_COMM_WORLD, &mpi_rank);
|
|
|
|
if(mpi_size == 1) return 0;
|
|
|
|
/* Create a parallel netcdf-4 file. */
|
|
/* nc_set_log_level(NC_TURN_OFF_LOGGING); */
|
|
/* nc_set_log_level(4);*/
|
|
|
|
if (nc_create_par(file_name, facc_type, comm, info, &ncid)) ERR;
|
|
|
|
/* The case is two dimensional variables, no unlimited dimension */
|
|
|
|
/* Create two dimensions. */
|
|
if (nc_def_dim(ncid, "d1", DIMSIZE2, dimids)) ERR;
|
|
if (nc_def_dim(ncid, "d2", DIMSIZE, &dimids[1])) ERR;
|
|
|
|
/* Create one var. */
|
|
if (nc_def_var(ncid, "v1", NC_INT, NDIMS1, dimids, &nvid)) ERR;
|
|
|
|
if (nc_enddef(ncid)) ERR;
|
|
|
|
|
|
/* hyperslab illustration for 3-processor case
|
|
|
|
--------
|
|
|aaaacccc|
|
|
|aaaacccc|
|
|
|bbbb |
|
|
|bbbb |
|
|
--------
|
|
*/
|
|
|
|
/* odd number of processors should be treated differently */
|
|
if(mpi_size%2 != 0) {
|
|
|
|
count_atom = DIMSIZE*2/(mpi_size+1);
|
|
if(mpi_rank <= mpi_size/2) {
|
|
start[0] = 0;
|
|
start[1] = mpi_rank*count_atom;
|
|
count[0] = DIMSIZE2/2;
|
|
count[1] = count_atom;
|
|
}
|
|
else {
|
|
start[0] = DIMSIZE2/2;
|
|
start[1] = (mpi_rank-mpi_size/2-1)*count_atom;
|
|
count[0] = DIMSIZE2/2;
|
|
count[1] = count_atom;
|
|
}
|
|
}
|
|
else {
|
|
|
|
count_atom = DIMSIZE*2/mpi_size;
|
|
if(mpi_rank < mpi_size/2) {
|
|
start[0] = 0;
|
|
start[1] = mpi_rank*count_atom;
|
|
count[0] = DIMSIZE2/2;
|
|
count[1] = count_atom;
|
|
}
|
|
else {
|
|
start[0] = DIMSIZE2/2;
|
|
start[1] = (mpi_rank-mpi_size/2)*count_atom;
|
|
count[0] = DIMSIZE2/2;
|
|
count[1] = count_atom;
|
|
}
|
|
}
|
|
|
|
if (nc_var_par_access(ncid, nvid, flag)) ERR;
|
|
data = malloc(sizeof(int)*count[1]*count[0]);
|
|
tempdata = data;
|
|
for (j=0; j<count[0];j++){
|
|
for (i=0; i<count[1]; i++){
|
|
*tempdata = mpi_rank*(j+1);
|
|
tempdata ++;
|
|
}
|
|
}
|
|
|
|
|
|
if (nc_put_vara_int(ncid, nvid, start, count, data)) ERR;
|
|
free(data);
|
|
|
|
/* Close the netcdf file. */
|
|
if (nc_close(ncid)) ERR;
|
|
|
|
if (nc_open_par(file_name, facc_type_open, comm, info, &ncid)) ERR;
|
|
|
|
/* Inquiry the variable */
|
|
if (nc_inq_varid(ncid, "v1", &rvid)) ERR;
|
|
|
|
if (nc_var_par_access(ncid, rvid, flag)) ERR;
|
|
|
|
rdata = malloc(sizeof(int)*count[1]*count[0]);
|
|
/* Read the data with the same slab settings */
|
|
if (nc_get_vara_int(ncid, rvid, start, count, rdata)) ERR;
|
|
|
|
temprdata = rdata;
|
|
for (j=0; j<count[0];j++){
|
|
for (i=0; i<count[1]; i++){
|
|
if(*temprdata != mpi_rank*(j+1))
|
|
{
|
|
res = -1;
|
|
break;
|
|
}
|
|
temprdata++;
|
|
}
|
|
}
|
|
|
|
free(rdata);
|
|
if(res == -1) ERR_RET;
|
|
|
|
/* Close the netcdf file. */
|
|
if (nc_close(ncid)) ERR;
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* test extending variables */
|
|
int test_pio_extend(int flag){
|
|
int rank, procs;
|
|
int ncFile;
|
|
int ncDimPart;
|
|
int ncDimVrtx;
|
|
int ncVarVrtx;
|
|
int dimsVrtx[2];
|
|
size_t start[2];
|
|
size_t count[2];
|
|
int vertices[] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
|
|
|
|
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
|
|
MPI_Comm_size(MPI_COMM_WORLD, &procs);
|
|
|
|
/* Create netcdf file */
|
|
if (nc_create_par("test.nc", NC_NETCDF4 | NC_MPIIO, MPI_COMM_WORLD, MPI_INFO_NULL, &ncFile)) ERR;
|
|
|
|
/* Create netcdf dimensions */
|
|
if (nc_def_dim(ncFile, "partitions", procs, &ncDimPart)) ERR;
|
|
if (nc_def_dim(ncFile, "vertices", NC_UNLIMITED, &ncDimVrtx)) ERR;
|
|
|
|
/* Create netcdf variables */
|
|
dimsVrtx[0] = ncDimPart;
|
|
dimsVrtx[1] = ncDimVrtx;
|
|
if (nc_def_var(ncFile, "vertex", NC_INT, 2, dimsVrtx, &ncVarVrtx)) ERR;
|
|
|
|
/* Start writing data */
|
|
if (nc_enddef(ncFile)) ERR;
|
|
|
|
/* Set access mode */
|
|
if (nc_var_par_access(ncFile, ncVarVrtx, flag)) ERR;
|
|
|
|
/* Write vertices */
|
|
start[0] = rank;
|
|
start[1] = 0;
|
|
count[0] = 1;
|
|
count[1] = rank;
|
|
if (nc_put_vara_int(ncFile, ncVarVrtx, start, count, vertices)) ERR;
|
|
|
|
/* Close netcdf file */
|
|
if (nc_close(ncFile)) ERR;
|
|
|
|
return 0;
|
|
}
|
|
|
|
/*-------------------------------------------------------------------------
|
|
* Function: getenv_all
|
|
*
|
|
* Purpose: Used to get the environment that the root MPI task has.
|
|
* name specifies which environment variable to look for
|
|
* val is the string to which the value of that environment
|
|
* variable will be copied.
|
|
*
|
|
* NOTE: The pointer returned by this function is only
|
|
* valid until the next call to getenv_all and the data
|
|
* stored there must be copied somewhere else before any
|
|
* further calls to getenv_all take place.
|
|
*
|
|
* Return: pointer to a string containing the value of the environment variable
|
|
* NULL if the varialbe doesn't exist in task 'root's environment.
|
|
*
|
|
* Programmer: Leon Arber
|
|
* 4/4/05
|
|
*
|
|
* Modifications:
|
|
*
|
|
*-------------------------------------------------------------------------
|
|
*/
|
|
|
|
char* getenv_all(MPI_Comm comm, int root, const char* name)
|
|
{
|
|
int nID;
|
|
int len = -1;
|
|
static char* env = NULL;
|
|
|
|
assert(name);
|
|
|
|
MPI_Comm_rank(comm, &nID);
|
|
|
|
/* The root task does the getenv call
|
|
* and sends the result to the other tasks */
|
|
if(nID == root)
|
|
{
|
|
env = getenv(name);
|
|
if(env)
|
|
{
|
|
len = strlen(env);
|
|
MPI_Bcast(&len, 1, MPI_INT, root, comm);
|
|
MPI_Bcast(env, len, MPI_CHAR, root, comm);
|
|
}
|
|
/* len -1 indicates that the variable was not in the environment */
|
|
else
|
|
MPI_Bcast(&len, 1, MPI_INT, root, comm);
|
|
}
|
|
else
|
|
{
|
|
MPI_Bcast(&len, 1, MPI_INT, root, comm);
|
|
if(len >= 0)
|
|
{
|
|
if(env == NULL)
|
|
env = (char*) malloc(len+1);
|
|
else if(strlen(env) < len)
|
|
env = (char*) realloc(env, len+1);
|
|
|
|
MPI_Bcast(env, len, MPI_CHAR, root, comm);
|
|
env[len] = '\0';
|
|
}
|
|
else
|
|
{
|
|
if(env)
|
|
free(env);
|
|
env = NULL;
|
|
}
|
|
}
|
|
|
|
MPI_Barrier(comm);
|
|
|
|
return env;
|
|
}
|
|
|
|
|