hdf5/testpar/testphdf5.c

620 lines
20 KiB
C
Raw Normal View History

/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* Copyright by The HDF Group. *
* 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 COPYING file, which can be found at the root of the source code *
* distribution tree, or in https://support.hdfgroup.org/ftp/HDF5/releases. *
* If you do not have access to either file, you may request a copy from *
* help@hdfgroup.org. *
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
1998-02-17 05:06:48 +08:00
/*
* Main driver of the Parallel HDF5 tests
1998-02-17 05:06:48 +08:00
*/
#include "testphdf5.h"
#ifndef PATH_MAX
#define PATH_MAX 512
#endif /* !PATH_MAX */
/* global variables */
int dim0;
int dim1;
int chunkdim0;
int chunkdim1;
int nerrors = 0; /* errors count */
int ndatasets = 300; /* number of datasets to create*/
int ngroups = 512; /* number of groups to create in root
* group. */
int facc_type = FACC_MPIO; /*Test file access type */
int dxfer_coll_type = DXFER_COLLECTIVE_IO;
H5E_auto2_t old_func; /* previous error handler */
void *old_client_data; /* previous error handler arg.*/
/* other option flags */
/* FILENAME and filenames must have the same number of names.
* Use PARATESTFILE in general and use a separated filename only if the file
* created in one test is accessed by a different test.
* filenames[0] is reserved as the file name for PARATESTFILE.
*/
#define NFILENAME 2
#define PARATESTFILE filenames[0]
const char *FILENAME[NFILENAME]={
"ParaTest",
NULL};
char filenames[NFILENAME][PATH_MAX];
hid_t fapl; /* file access property list */
#ifdef USE_PAUSE
/* pause the process for a moment to allow debugger to attach if desired. */
/* Will pause more if greenlight file is not persent but will eventually */
/* continue. */
#include <sys/types.h>
#include <sys/stat.h>
void pause_proc(void)
{
int pid;
h5_stat_t statbuf;
char greenlight[] = "go";
int maxloop = 10;
int loops = 0;
int time_int = 10;
/* mpi variables */
int mpi_size, mpi_rank;
int mpi_namelen;
char mpi_name[MPI_MAX_PROCESSOR_NAME];
pid = getpid();
MPI_Comm_size(MPI_COMM_WORLD, &mpi_size);
MPI_Comm_rank(MPI_COMM_WORLD, &mpi_rank);
MPI_Get_processor_name(mpi_name, &mpi_namelen);
if (MAINPROCESS)
while ((HDstat(greenlight, &statbuf) == -1) && loops < maxloop){
if (!loops++){
printf("Proc %d (%*s, %d): to debug, attach %d\n",
mpi_rank, mpi_namelen, mpi_name, pid, pid);
}
printf("waiting(%ds) for file %s ...\n", time_int, greenlight);
fflush(stdout);
HDsleep(time_int);
}
MPI_Barrier(MPI_COMM_WORLD);
}
/* Use the Profile feature of MPI to call the pause_proc() */
int MPI_Init(int *argc, char ***argv)
{
int ret_code;
ret_code=PMPI_Init(argc, argv);
pause_proc();
return (ret_code);
}
#endif /* USE_PAUSE */
/*
* Show command usage
*/
static void
usage(void)
{
printf(" [-r] [-w] [-m<n_datasets>] [-n<n_groups>] "
"[-o] [-f <prefix>] [-d <dim0> <dim1>]\n");
printf("\t-m<n_datasets>"
"\tset number of datasets for the multiple dataset test\n");
printf("\t-n<n_groups>"
"\tset number of groups for the multiple group test\n");
printf("\t-f <prefix>\tfilename prefix\n");
printf("\t-2\t\tuse Split-file together with MPIO\n");
printf("\t-d <factor0> <factor1>\tdataset dimensions factors. Defaults (%d,%d)\n",
ROW_FACTOR, COL_FACTOR);
printf("\t-c <dim0> <dim1>\tdataset chunk dimensions. Defaults (dim0/10,dim1/10)\n");
printf("\n");
}
/*
* parse the command line options
*/
static int
parse_options(int argc, char **argv)
{
int mpi_size, mpi_rank; /* mpi variables */
MPI_Comm_size(MPI_COMM_WORLD, &mpi_size);
MPI_Comm_rank(MPI_COMM_WORLD, &mpi_rank);
/* setup default chunk-size. Make sure sizes are > 0 */
chunkdim0 = (dim0+9)/10;
chunkdim1 = (dim1+9)/10;
while (--argc){
if (**(++argv) != '-'){
break;
}else{
switch(*(*argv+1)){
case 'm': ndatasets = atoi((*argv+1)+1);
if (ndatasets < 0){
nerrors++;
return(1);
}
break;
case 'n': ngroups = atoi((*argv+1)+1);
if (ngroups < 0){
nerrors++;
return(1);
}
break;
case 'f': if (--argc < 1) {
nerrors++;
return(1);
}
if (**(++argv) == '-') {
nerrors++;
return(1);
}
paraprefix = *argv;
break;
case 'i': /* Collective MPI-IO access with independent IO */
dxfer_coll_type = DXFER_INDEPENDENT_IO;
break;
case '2': /* Use the split-file driver with MPIO access */
/* Can use $HDF5_METAPREFIX to define the */
/* meta-file-prefix. */
facc_type = FACC_MPIO | FACC_SPLIT;
break;
case 'd': /* dimensizes */
if (--argc < 2){
nerrors++;
return(1);
}
dim0 = atoi(*(++argv))*mpi_size;
argc--;
dim1 = atoi(*(++argv))*mpi_size;
/* set default chunkdim sizes too */
chunkdim0 = (dim0+9)/10;
chunkdim1 = (dim1+9)/10;
break;
case 'c': /* chunk dimensions */
if (--argc < 2){
nerrors++;
return(1);
}
chunkdim0 = atoi(*(++argv));
argc--;
chunkdim1 = atoi(*(++argv));
break;
case 'h': /* print help message--return with nerrors set */
return(1);
default: printf("Illegal option(%s)\n", *argv);
nerrors++;
return(1);
}
}
} /*while*/
/* check validity of dimension and chunk sizes */
if (dim0 <= 0 || dim1 <= 0){
printf("Illegal dim sizes (%d, %d)\n", dim0, dim1);
nerrors++;
return(1);
}
if (chunkdim0 <= 0 || chunkdim1 <= 0){
printf("Illegal chunkdim sizes (%d, %d)\n", chunkdim0, chunkdim1);
nerrors++;
return(1);
}
/* Make sure datasets can be divided into equal portions by the processes */
if ((dim0 % mpi_size) || (dim1 % mpi_size)){
if (MAINPROCESS)
printf("dim0(%d) and dim1(%d) must be multiples of processes(%d)\n",
dim0, dim1, mpi_size);
nerrors++;
return(1);
}
/* compose the test filenames */
{
int i, n;
n = sizeof(FILENAME)/sizeof(FILENAME[0]) - 1; /* exclude the NULL */
for (i=0; i < n; i++)
if (h5_fixname(FILENAME[i],fapl,filenames[i],sizeof(filenames[i]))
== NULL){
printf("h5_fixname failed\n");
nerrors++;
return(1);
}
printf("Test filenames are:\n");
for (i=0; i < n; i++)
printf(" %s\n", filenames[i]);
}
return(0);
}
/*
* Create the appropriate File access property list
*/
hid_t
create_faccess_plist(MPI_Comm comm, MPI_Info info, int l_facc_type)
{
hid_t ret_pl = -1;
herr_t ret; /* generic return value */
int mpi_rank; /* mpi variables */
/* need the rank for error checking macros */
MPI_Comm_rank(MPI_COMM_WORLD, &mpi_rank);
ret_pl = H5Pcreate (H5P_FILE_ACCESS);
VRFY((ret_pl >= 0), "H5P_FILE_ACCESS");
if (l_facc_type == FACC_DEFAULT)
return (ret_pl);
if (l_facc_type == FACC_MPIO){
/* set Parallel access with communicator */
ret = H5Pset_fapl_mpio(ret_pl, comm, info);
VRFY((ret >= 0), "");
ret = H5Pset_all_coll_metadata_ops(ret_pl, TRUE);
VRFY((ret >= 0), "");
ret = H5Pset_coll_metadata_write(ret_pl, TRUE);
VRFY((ret >= 0), "");
return(ret_pl);
}
if (l_facc_type == (FACC_MPIO | FACC_SPLIT)){
hid_t mpio_pl;
mpio_pl = H5Pcreate (H5P_FILE_ACCESS);
VRFY((mpio_pl >= 0), "");
/* set Parallel access with communicator */
ret = H5Pset_fapl_mpio(mpio_pl, comm, info);
VRFY((ret >= 0), "");
/* setup file access template */
ret_pl = H5Pcreate (H5P_FILE_ACCESS);
VRFY((ret_pl >= 0), "");
/* set Parallel access with communicator */
ret = H5Pset_fapl_split(ret_pl, ".meta", mpio_pl, ".raw", mpio_pl);
VRFY((ret >= 0), "H5Pset_fapl_split succeeded");
H5Pclose(mpio_pl);
return(ret_pl);
}
/* unknown file access types */
return (ret_pl);
}
int main(int argc, char **argv)
{
int mpi_size, mpi_rank; /* mpi variables */
H5Ptest_param_t ndsets_params, ngroups_params;
H5Ptest_param_t collngroups_params;
H5Ptest_param_t io_mode_confusion_params;
H5Ptest_param_t rr_obj_flush_confusion_params;
#ifndef H5_HAVE_WIN32_API
/* Un-buffer the stdout and stderr */
HDsetbuf(stderr, NULL);
HDsetbuf(stdout, NULL);
#endif
MPI_Init(&argc, &argv);
MPI_Comm_size(MPI_COMM_WORLD, &mpi_size);
MPI_Comm_rank(MPI_COMM_WORLD, &mpi_rank);
dim0 = ROW_FACTOR*mpi_size;
dim1 = COL_FACTOR*mpi_size;
if (MAINPROCESS){
printf("===================================\n");
printf("PHDF5 TESTS START\n");
printf("===================================\n");
}
/* Attempt to turn off atexit post processing so that in case errors
* happen during the test and the process is aborted, it will not get
* hang in the atexit post processing in which it may try to make MPI
* calls. By then, MPI calls may not work.
*/
if (H5dont_atexit() < 0){
printf("Failed to turn off atexit processing. Continue.\n");
};
[svn-r6709] Purpose: Bug fixes/API changes Description: Previously, the Communicator and Info object arguments supplied to H5Pset_fapl_mpio() are stored in the property with its handle values. This meant changes to the communicator or the Info object after calling H5Pset_fapl_mpio would affect the how the property list function. This was also the case when H5Fopen/create operated. They just stored the handle value. This is not according to the MPI-2 defined behavior of how Info objects should be handled. (MPI-2 defines Info objects must be parsed when called.) The old design was trying to avoid numerous duplicates of the same information (e.g., every property object holds one version, every file opened holds another version, when all of them are referring to the same original version.) Nevertheless it is safer to implement it according to MPI-2 definition. Futhermore, the library often needs to do message passing using the supplied communicator. Using the same communicator as the application version may result in some messages mix up. Solution: H5Pset_fapl_mpio now stores a duplicate of each of the communicator and Info object. H5Pget_fapl_mpio returns a duplicate of its stored communicator and Info object. It is now the responsibility of the applications to free those objects when done. H5Fopen/create also stores a duplicate of the communicator and Info object supplied by the File Access Property list. H5Fclose frees those duplicates. There are a few more internal VFL call back functions that they follow this "make duplicates" requirement. Platforms tested: "h5committested". What other platforms/configurations were tested? Eirene (mpicc), sol(mpicc), copper(parallel) Misc. update: Need to update MANIFEST for the added t_ph5basic.c which tests the correctness of duplicated communicator and INFO object.
2003-04-18 11:04:56 +08:00
H5open();
h5_show_hostname();
/* Initialize testing framework */
TestInit(argv[0], usage, parse_options);
/* Tests are generally arranged from least to most complexity... */
AddTest("mpiodup", test_fapl_mpio_dup, NULL,
"fapl_mpio duplicate", NULL);
AddTest("split", test_split_comm_access, NULL,
"dataset using split communicators", PARATESTFILE);
#ifdef PB_OUT /* temporary: disable page buffering when parallel */
AddTest("page_buffer", test_page_buffer_access, NULL,
"page buffer usage in parallel", PARATESTFILE);
#endif
AddTest("props", test_file_properties, NULL,
"Coll Metadata file property settings", PARATESTFILE);
AddTest("idsetw", dataset_writeInd, NULL,
"dataset independent write", PARATESTFILE);
AddTest("idsetr", dataset_readInd, NULL,
"dataset independent read", PARATESTFILE);
AddTest("cdsetw", dataset_writeAll, NULL,
"dataset collective write", PARATESTFILE);
AddTest("cdsetr", dataset_readAll, NULL,
"dataset collective read", PARATESTFILE);
AddTest("eidsetw", extend_writeInd, NULL,
"extendible dataset independent write", PARATESTFILE);
AddTest("eidsetr", extend_readInd, NULL,
"extendible dataset independent read", PARATESTFILE);
AddTest("ecdsetw", extend_writeAll, NULL,
"extendible dataset collective write", PARATESTFILE);
AddTest("ecdsetr", extend_readAll, NULL,
"extendible dataset collective read", PARATESTFILE);
AddTest("eidsetw2", extend_writeInd2, NULL,
"extendible dataset independent write #2", PARATESTFILE);
AddTest("selnone", none_selection_chunk, NULL,
"chunked dataset with none-selection", PARATESTFILE);
AddTest("calloc", test_chunk_alloc, NULL,
"parallel extend Chunked allocation on serial file", PARATESTFILE);
AddTest("fltread", test_filter_read, NULL,
"parallel read of dataset written serially with filters", PARATESTFILE);
#ifdef H5_HAVE_FILTER_DEFLATE
AddTest("cmpdsetr", compress_readAll, NULL,
"compressed dataset collective read", PARATESTFILE);
#endif /* H5_HAVE_FILTER_DEFLATE */
AddTest("zerodsetr", zero_dim_dset, NULL,
"zero dim dset", PARATESTFILE);
ndsets_params.name = PARATESTFILE;
ndsets_params.count = ndatasets;
AddTest("ndsetw", multiple_dset_write, NULL,
"multiple datasets write", &ndsets_params);
ngroups_params.name = PARATESTFILE;
ngroups_params.count = ngroups;
AddTest("ngrpw", multiple_group_write, NULL,
"multiple groups write", &ngroups_params);
AddTest("ngrpr", multiple_group_read, NULL,
"multiple groups read", &ngroups_params);
AddTest("compact", compact_dataset, NULL,
"compact dataset test", PARATESTFILE);
collngroups_params.name = PARATESTFILE;
collngroups_params.count = ngroups;
AddTest("cngrpw", collective_group_write, NULL,
"collective group and dataset write", &collngroups_params);
AddTest("ingrpr", independent_group_read, NULL,
"independent group and dataset read", &collngroups_params);
#ifndef H5_HAVE_WIN32_API
AddTest("bigdset", big_dataset, NULL,
"big dataset test", PARATESTFILE);
#else
printf("big dataset test will be skipped on Windows (JIRA HDDFV-8064)\n");
#endif
AddTest("fill", dataset_fillvalue, NULL,
"dataset fill value", PARATESTFILE);
AddTest("cchunk1",
coll_chunk1,NULL, "simple collective chunk io",PARATESTFILE);
AddTest("cchunk2",
coll_chunk2,NULL, "noncontiguous collective chunk io",PARATESTFILE);
AddTest("cchunk3",
coll_chunk3,NULL, "multi-chunk collective chunk io",PARATESTFILE);
AddTest("cchunk4",
coll_chunk4,NULL, "collective chunk io with partial non-selection ",PARATESTFILE);
if((mpi_size < 3)&& MAINPROCESS ) {
printf("Collective chunk IO optimization APIs ");
printf("needs at least 3 processes to participate\n");
printf("Collective chunk IO API tests will be skipped \n");
}
AddTest((mpi_size <3)? "-cchunk5":"cchunk5" ,
coll_chunk5,NULL,
"linked chunk collective IO without optimization",PARATESTFILE);
AddTest((mpi_size < 3)? "-cchunk6" : "cchunk6",
coll_chunk6,NULL,
"multi-chunk collective IO with direct request",PARATESTFILE);
AddTest((mpi_size < 3)? "-cchunk7" : "cchunk7",
coll_chunk7,NULL,
"linked chunk collective IO with optimization",PARATESTFILE);
AddTest((mpi_size < 3)? "-cchunk8" : "cchunk8",
coll_chunk8,NULL,
"linked chunk collective IO transferring to multi-chunk",PARATESTFILE);
AddTest((mpi_size < 3)? "-cchunk9" : "cchunk9",
coll_chunk9,NULL,
"multiple chunk collective IO with optimization",PARATESTFILE);
AddTest((mpi_size < 3)? "-cchunk10" : "cchunk10",
coll_chunk10,NULL,
"multiple chunk collective IO transferring to independent IO",PARATESTFILE);
/* irregular collective IO tests*/
AddTest("ccontw",
coll_irregular_cont_write,NULL,
"collective irregular contiguous write",PARATESTFILE);
AddTest("ccontr",
coll_irregular_cont_read,NULL,
"collective irregular contiguous read",PARATESTFILE);
AddTest("cschunkw",
coll_irregular_simple_chunk_write,NULL,
"collective irregular simple chunk write",PARATESTFILE);
AddTest("cschunkr",
coll_irregular_simple_chunk_read,NULL,
"collective irregular simple chunk read",PARATESTFILE);
AddTest("ccchunkw",
coll_irregular_complex_chunk_write,NULL,
"collective irregular complex chunk write",PARATESTFILE);
AddTest("ccchunkr",
coll_irregular_complex_chunk_read,NULL,
"collective irregular complex chunk read",PARATESTFILE);
AddTest("null", null_dataset, NULL,
"null dataset test", PARATESTFILE);
io_mode_confusion_params.name = PARATESTFILE;
io_mode_confusion_params.count = 0; /* value not used */
AddTest("I/Omodeconf", io_mode_confusion, NULL,
"I/O mode confusion test -- hangs quickly on failure",
&io_mode_confusion_params);
if((mpi_size < 3) && MAINPROCESS) {
printf("rr_obj_hdr_flush_confusion test needs at least 3 processes.\n");
printf("rr_obj_hdr_flush_confusion test will be skipped \n");
}
if(mpi_size > 2) {
rr_obj_flush_confusion_params.name = PARATESTFILE;
rr_obj_flush_confusion_params.count = 0; /* value not used */
AddTest("rrobjflushconf", rr_obj_hdr_flush_confusion, NULL,
"round robin object header flush confusion test",
&rr_obj_flush_confusion_params);
}
AddTest("alnbg1",
chunk_align_bug_1, NULL,
"Chunk allocation with alignment bug.",
PARATESTFILE);
AddTest("tldsc",
lower_dim_size_comp_test, NULL,
"test lower dim size comp in span tree to mpi derived type",
PARATESTFILE);
AddTest("lccio",
link_chunk_collective_io_test, NULL,
"test mpi derived type management",
PARATESTFILE);
AddTest("actualio", actual_io_mode_tests, NULL,
"test actual io mode proprerty",
PARATESTFILE);
AddTest("nocolcause", no_collective_cause_tests, NULL,
"test cause for broken collective io",
PARATESTFILE);
AddTest("edpl", test_plist_ed, NULL,
"encode/decode Property Lists", NULL);
if((mpi_size < 2) && MAINPROCESS) {
printf("File Image Ops daisy chain test needs at least 2 processes.\n");
printf("File Image Ops daisy chain test will be skipped \n");
}
AddTest((mpi_size < 2)? "-fiodc" : "fiodc", file_image_daisy_chain_test, NULL,
"file image ops daisy chain", NULL);
if((mpi_size < 2)&& MAINPROCESS ) {
printf("Atomicity tests need at least 2 processes to participate\n");
printf("8 is more recommended.. Atomicity tests will be skipped \n");
}
else if (facc_type != FACC_MPIO && MAINPROCESS) {
printf("Atomicity tests will not work with a non MPIO VFD\n");
}
else if(mpi_size >= 2 && facc_type == FACC_MPIO){
AddTest("atomicity", dataset_atomicity, NULL,
"dataset atomic updates", PARATESTFILE);
}
AddTest("denseattr", test_dense_attr, NULL,
"Store Dense Attributes", PARATESTFILE);
AddTest("noselcollmdread", test_partial_no_selection_coll_md_read, NULL,
"Collective Metadata read with some ranks having no selection", PARATESTFILE);
AddTest("MC coll MD read", test_multi_chunk_io_addrmap_issue, NULL,
"Collective MD read with multi chunk I/O (H5D__chunk_addrmap)", PARATESTFILE);
AddTest("LC coll MD read", test_link_chunk_io_sort_chunk_issue, NULL,
"Collective MD read with link chunk I/O (H5D__sort_chunk)", PARATESTFILE);
/* Display testing information */
TestInfo(argv[0]);
/* setup file access property list */
fapl = H5Pcreate (H5P_FILE_ACCESS);
H5Pset_fapl_mpio(fapl, MPI_COMM_WORLD, MPI_INFO_NULL);
/* Parse command line arguments */
TestParseCmdLine(argc, argv);
if (dxfer_coll_type == DXFER_INDEPENDENT_IO && MAINPROCESS){
printf("===================================\n"
" Using Independent I/O with file set view to replace collective I/O \n"
"===================================\n");
}
/* Perform requested testing */
PerformTests();
/* make sure all processes are finished before final report, cleanup
* and exit.
*/
MPI_Barrier(MPI_COMM_WORLD);
/* Display test summary, if requested */
if (MAINPROCESS && GetTestSummary())
TestSummary();
/* Clean up test files */
h5_clean_files(FILENAME, fapl);
nerrors += GetTestNumErrs();
/* Gather errors from all processes */
{
int temp;
MPI_Allreduce(&nerrors, &temp, 1, MPI_INT, MPI_MAX, MPI_COMM_WORLD);
nerrors=temp;
}
if (MAINPROCESS){ /* only process 0 reports */
printf("===================================\n");
if (nerrors)
printf("***PHDF5 tests detected %d errors***\n", nerrors);
else
printf("PHDF5 tests finished with no errors\n");
printf("===================================\n");
}
/* close HDF5 library */
H5close();
[svn-r6709] Purpose: Bug fixes/API changes Description: Previously, the Communicator and Info object arguments supplied to H5Pset_fapl_mpio() are stored in the property with its handle values. This meant changes to the communicator or the Info object after calling H5Pset_fapl_mpio would affect the how the property list function. This was also the case when H5Fopen/create operated. They just stored the handle value. This is not according to the MPI-2 defined behavior of how Info objects should be handled. (MPI-2 defines Info objects must be parsed when called.) The old design was trying to avoid numerous duplicates of the same information (e.g., every property object holds one version, every file opened holds another version, when all of them are referring to the same original version.) Nevertheless it is safer to implement it according to MPI-2 definition. Futhermore, the library often needs to do message passing using the supplied communicator. Using the same communicator as the application version may result in some messages mix up. Solution: H5Pset_fapl_mpio now stores a duplicate of each of the communicator and Info object. H5Pget_fapl_mpio returns a duplicate of its stored communicator and Info object. It is now the responsibility of the applications to free those objects when done. H5Fopen/create also stores a duplicate of the communicator and Info object supplied by the File Access Property list. H5Fclose frees those duplicates. There are a few more internal VFL call back functions that they follow this "make duplicates" requirement. Platforms tested: "h5committested". What other platforms/configurations were tested? Eirene (mpicc), sol(mpicc), copper(parallel) Misc. update: Need to update MANIFEST for the added t_ph5basic.c which tests the correctness of duplicated communicator and INFO object.
2003-04-18 11:04:56 +08:00
/* Release test infrastructure */
TestShutdown();
/* MPI_Finalize must be called AFTER H5close which may use MPI calls */
MPI_Finalize();
/* cannot just return (nerrors) because exit code is limited to 1byte */
return(nerrors!=0);
}