diff options
Diffstat (limited to 'testpar')
| -rw-r--r-- | testpar/t_filters_parallel.c | 231 | ||||
| -rw-r--r-- | testpar/t_filters_parallel.h | 2 |
2 files changed, 152 insertions, 81 deletions
diff --git a/testpar/t_filters_parallel.c b/testpar/t_filters_parallel.c index bdfde09..7dfb8bc 100644 --- a/testpar/t_filters_parallel.c +++ b/testpar/t_filters_parallel.c @@ -576,11 +576,21 @@ create_datasets(hid_t parent_obj_id, const char *dset_name, hid_t type_id, hid_t case USE_MULTIPLE_DATASETS: case USE_MULTIPLE_DATASETS_MIXED_FILTERED: dset_name_ptr = dset_name_multi_buf; - n_dsets = (rand() % (MAX_NUM_DSETS_MULTI - 1)) + 2; + + if (MAINPROCESS) + n_dsets = (rand() % (MAX_NUM_DSETS_MULTI - 1)) + 2; + + if (mpi_size > 1) + VRFY((MPI_SUCCESS == MPI_Bcast(&n_dsets, 1, MPI_INT, 0, comm)), "MPI_Bcast succeeded"); /* Select between 1 and (n_dsets - 1) datasets to be unfiltered */ if (test_mode == USE_MULTIPLE_DATASETS_MIXED_FILTERED) { - n_unfiltered = (rand() % (n_dsets - 1)) + 1; + if (MAINPROCESS) + n_unfiltered = (rand() % (n_dsets - 1)) + 1; + + if (mpi_size > 1) + VRFY((MPI_SUCCESS == MPI_Bcast(&n_unfiltered, 1, MPI_INT, 0, comm)), + "MPI_Bcast succeeded"); unfiltered_dcpl = H5Pcopy(dcpl_id); VRFY((unfiltered_dcpl >= 0), "H5Pcopy succeeded"); @@ -621,7 +631,11 @@ create_datasets(hid_t parent_obj_id, const char *dset_name, hid_t type_id, hid_t * remaining datasets as unfiltered datasets. Otherwise, * randomly determine if a dataset will be unfiltered. */ - unfiltered = ((size_t)n_unfiltered == dsets_left) || ((rand() % 2) == 0); + if (MAINPROCESS) + unfiltered = ((size_t)n_unfiltered == dsets_left) || ((rand() % 2) == 0); + + if (mpi_size > 1) + VRFY((MPI_SUCCESS == MPI_Bcast(&unfiltered, 1, MPI_C_BOOL, 0, comm)), "MPI_Bcast succeeded"); if (unfiltered) { curr_dcpl = unfiltered_dcpl; @@ -1201,6 +1215,12 @@ test_write_filtered_dataset_no_overlap_partial(const char *parent_group, H5Z_fil plist_id = H5Pcopy(dcpl_id); VRFY((plist_id >= 0), "DCPL copy succeeded"); + /* + * Since we're only doing a partial write to the dataset, make + * sure the fill time is set appropriately + */ + VRFY((H5Pset_fill_time(plist_id, H5D_FILL_TIME_IFSET) >= 0), "H5Pset_fill_time succeeded"); + VRFY((H5Pset_chunk(plist_id, WRITE_UNSHARED_FILTERED_CHUNKS_PARTIAL_DATASET_DIMS, chunk_dims) >= 0), "Chunk size set"); @@ -1485,6 +1505,7 @@ test_write_filtered_dataset_single_unlim_dim_no_overlap(const char *parent_group hsize_t block[WRITE_UNSHARED_ONE_UNLIM_DIM_DATASET_DIMS]; size_t data_size; size_t num_dsets; + size_t num_loops; hid_t dset_ids[MAX_NUM_DSETS_MULTI]; hid_t fspace_ids[MAX_NUM_DSETS_MULTI]; hid_t file_id = H5I_INVALID_HID, plist_id = H5I_INVALID_HID; @@ -1551,7 +1572,12 @@ test_write_filtered_dataset_single_unlim_dim_no_overlap(const char *parent_group read_bufs[dset_idx] = tmp_buf; } - for (size_t i = 0; i < (size_t)WRITE_UNSHARED_ONE_UNLIM_DIM_NLOOPS; i++) { + /* Determine number of loops to run through */ + num_loops = WRITE_UNSHARED_ONE_UNLIM_DIM_NLOOPS; + if ((test_mode == USE_MULTIPLE_DATASETS) || (test_mode == USE_MULTIPLE_DATASETS_MIXED_FILTERED)) + num_loops /= 2; + + for (size_t i = 0; i < num_loops; i++) { /* Each process defines the dataset selection in memory and writes * it to the hyperslab in the file */ @@ -1590,7 +1616,7 @@ test_write_filtered_dataset_single_unlim_dim_no_overlap(const char *parent_group VRFY((0 == memcmp(read_bufs[dset_idx], data_bufs[dset_idx], data_size)), "Data verification succeeded"); - if (i < (size_t)WRITE_UNSHARED_ONE_UNLIM_DIM_NLOOPS - 1) { + if (i < num_loops - 1) { /* Extend the dataset(s) by count[1] chunks in the extensible dimension */ dataset_dims[1] += count[1] * block[1]; @@ -1646,6 +1672,7 @@ test_write_filtered_dataset_single_unlim_dim_overlap(const char *parent_group, H hsize_t block[WRITE_SHARED_ONE_UNLIM_DIM_DATASET_DIMS]; size_t data_size; size_t num_dsets; + size_t num_loops; hid_t dset_ids[MAX_NUM_DSETS_MULTI]; hid_t fspace_ids[MAX_NUM_DSETS_MULTI]; hid_t file_id = H5I_INVALID_HID, plist_id = H5I_INVALID_HID; @@ -1712,7 +1739,12 @@ test_write_filtered_dataset_single_unlim_dim_overlap(const char *parent_group, H read_bufs[dset_idx] = tmp_buf; } - for (size_t i = 0; i < (size_t)WRITE_SHARED_ONE_UNLIM_DIM_NLOOPS; i++) { + /* Determine number of loops to run through */ + num_loops = WRITE_SHARED_ONE_UNLIM_DIM_NLOOPS; + if ((test_mode == USE_MULTIPLE_DATASETS) || (test_mode == USE_MULTIPLE_DATASETS_MIXED_FILTERED)) + num_loops /= 2; + + for (size_t i = 0; i < num_loops; i++) { /* Each process defines the dataset selection in memory and writes * it to the hyperslab in the file */ @@ -1750,7 +1782,7 @@ test_write_filtered_dataset_single_unlim_dim_overlap(const char *parent_group, H VRFY((0 == memcmp(read_bufs[dset_idx], data_bufs[dset_idx], data_size)), "Data verification succeeded"); - if (i < (size_t)WRITE_SHARED_ONE_UNLIM_DIM_NLOOPS - 1) { + if (i < num_loops - 1) { /* Extend the dataset(s) by count[1] chunks in the extensible dimension */ dataset_dims[1] += count[1] * block[1]; @@ -1808,6 +1840,7 @@ test_write_filtered_dataset_multi_unlim_dim_no_overlap(const char *parent_group, hsize_t block[WRITE_UNSHARED_TWO_UNLIM_DIM_DATASET_DIMS]; size_t data_size; size_t num_dsets; + size_t num_loops; hid_t dset_ids[MAX_NUM_DSETS_MULTI]; hid_t fspace_ids[MAX_NUM_DSETS_MULTI]; hid_t file_id = H5I_INVALID_HID, plist_id = H5I_INVALID_HID; @@ -1855,7 +1888,12 @@ test_write_filtered_dataset_multi_unlim_dim_no_overlap(const char *parent_group, VRFY((H5Sclose(filespace) >= 0), "File dataspace close succeeded"); - for (size_t i = 0; i < (size_t)WRITE_UNSHARED_TWO_UNLIM_DIM_NLOOPS; i++) { + /* Determine number of loops to run through */ + num_loops = WRITE_UNSHARED_TWO_UNLIM_DIM_NLOOPS; + if ((test_mode == USE_MULTIPLE_DATASETS) || (test_mode == USE_MULTIPLE_DATASETS_MIXED_FILTERED)) + num_loops /= 2; + + for (size_t i = 0; i < num_loops; i++) { /* Set selected dimensions */ sel_dims[0] = (i + 1) * WRITE_UNSHARED_TWO_UNLIM_DIM_CH_NROWS; sel_dims[1] = (i + 1) * WRITE_UNSHARED_TWO_UNLIM_DIM_CH_NCOLS; @@ -1916,7 +1954,7 @@ test_write_filtered_dataset_multi_unlim_dim_no_overlap(const char *parent_group, VRFY((0 == memcmp(read_bufs[dset_idx], data_bufs[dset_idx], data_size)), "Data verification succeeded"); - if (i < (size_t)WRITE_UNSHARED_TWO_UNLIM_DIM_NLOOPS - 1) { + if (i < num_loops - 1) { /* * Extend the dataset(s) by the size of one chunk per rank * in the first extensible dimension. Extend the dataset(s) @@ -1977,6 +2015,7 @@ test_write_filtered_dataset_multi_unlim_dim_overlap(const char *parent_group, H5 hsize_t block[WRITE_SHARED_TWO_UNLIM_DIM_DATASET_DIMS]; size_t data_size; size_t num_dsets; + size_t num_loops; hid_t dset_ids[MAX_NUM_DSETS_MULTI]; hid_t fspace_ids[MAX_NUM_DSETS_MULTI]; hid_t file_id = H5I_INVALID_HID, plist_id = H5I_INVALID_HID; @@ -2024,7 +2063,12 @@ test_write_filtered_dataset_multi_unlim_dim_overlap(const char *parent_group, H5 VRFY((H5Sclose(filespace) >= 0), "File dataspace close succeeded"); - for (size_t i = 0; i < (size_t)WRITE_SHARED_TWO_UNLIM_DIM_NLOOPS; i++) { + /* Determine number of loops to run through */ + num_loops = WRITE_SHARED_TWO_UNLIM_DIM_NLOOPS; + if ((test_mode == USE_MULTIPLE_DATASETS) || (test_mode == USE_MULTIPLE_DATASETS_MIXED_FILTERED)) + num_loops /= 2; + + for (size_t i = 0; i < num_loops; i++) { /* Set selected dimensions */ sel_dims[0] = (i + 1); sel_dims[1] = (i + 1) * (size_t)WRITE_SHARED_TWO_UNLIM_DIM_CH_NCOLS; @@ -2085,7 +2129,7 @@ test_write_filtered_dataset_multi_unlim_dim_overlap(const char *parent_group, H5 VRFY((0 == memcmp(read_bufs[dset_idx], data_bufs[dset_idx], data_size)), "Data verification succeeded"); - if (i < (size_t)WRITE_SHARED_TWO_UNLIM_DIM_NLOOPS - 1) { + if (i < num_loops - 1) { /* Extend the dataset(s) by the size of a chunk in each extensible dimension */ dataset_dims[0] += (hsize_t)WRITE_SHARED_TWO_UNLIM_DIM_CH_NROWS; dataset_dims[1] += (hsize_t)WRITE_SHARED_TWO_UNLIM_DIM_CH_NCOLS; @@ -2177,6 +2221,12 @@ test_write_filtered_dataset_single_no_selection(const char *parent_group, H5Z_fi plist_id = H5Pcopy(dcpl_id); VRFY((plist_id >= 0), "DCPL copy succeeded"); + /* + * Since we're only doing a partial write to the dataset, make + * sure the fill time is set appropriately + */ + VRFY((H5Pset_fill_time(plist_id, H5D_FILL_TIME_IFSET) >= 0), "H5Pset_fill_time succeeded"); + VRFY((H5Pset_chunk(plist_id, WRITE_SINGLE_NO_SELECTION_FILTERED_CHUNKS_DATASET_DIMS, chunk_dims) >= 0), "Chunk size set"); @@ -2341,6 +2391,12 @@ test_write_filtered_dataset_all_no_selection(const char *parent_group, H5Z_filte plist_id = H5Pcopy(dcpl_id); VRFY((plist_id >= 0), "DCPL copy succeeded"); + /* + * Since we're doing a no-op write to the dataset, + * make sure the fill time is set appropriately + */ + VRFY((H5Pset_fill_time(plist_id, H5D_FILL_TIME_IFSET) >= 0), "H5Pset_fill_time succeeded"); + VRFY((H5Pset_chunk(plist_id, WRITE_ALL_NO_SELECTION_FILTERED_CHUNKS_DATASET_DIMS, chunk_dims) >= 0), "Chunk size set"); @@ -5250,7 +5306,6 @@ test_read_filtered_dataset_all_no_selection(const char *parent_group, H5Z_filter void *read_bufs[MAX_NUM_DSETS_MULTI] = {0}; hsize_t dataset_dims[READ_ALL_NO_SELECTION_FILTERED_CHUNKS_DATASET_DIMS]; hsize_t chunk_dims[READ_ALL_NO_SELECTION_FILTERED_CHUNKS_DATASET_DIMS]; - hsize_t sel_dims[READ_ALL_NO_SELECTION_FILTERED_CHUNKS_DATASET_DIMS]; size_t data_size, read_buf_size; size_t num_dsets; hid_t dset_ids[MAX_NUM_DSETS_MULTI]; @@ -5368,8 +5423,6 @@ test_read_filtered_dataset_all_no_selection(const char *parent_group, H5Z_filter open_datasets(group_id, READ_ALL_NO_SELECTION_FILTERED_CHUNKS_DATASET_NAME, num_dsets, test_mode, dset_ids); - sel_dims[0] = sel_dims[1] = 0; - select_none(num_dsets, dset_ids, fspace_ids); read_buf_size = dataset_dims[0] * dataset_dims[1] * sizeof(C_DATATYPE); @@ -8275,6 +8328,7 @@ test_shrinking_growing_chunks(const char *parent_group, H5Z_filter_t filter_id, hsize_t block[SHRINKING_GROWING_CHUNKS_DATASET_DIMS]; size_t data_size; size_t num_dsets; + size_t num_loops; hid_t dset_ids[MAX_NUM_DSETS_MULTI]; hid_t fspace_ids[MAX_NUM_DSETS_MULTI]; hid_t file_id = H5I_INVALID_HID, plist_id = H5I_INVALID_HID; @@ -8352,7 +8406,12 @@ test_shrinking_growing_chunks(const char *parent_group, H5Z_filter_t filter_id, read_bufs[dset_idx] = tmp_buf; } - for (size_t i = 0; i < SHRINKING_GROWING_CHUNKS_NLOOPS; i++) { + /* Determine number of loops to run through */ + num_loops = SHRINKING_GROWING_CHUNKS_NLOOPS; + if ((test_mode == USE_MULTIPLE_DATASETS) || (test_mode == USE_MULTIPLE_DATASETS_MIXED_FILTERED)) + num_loops /= 2; + + for (size_t i = 0; i < num_loops; i++) { for (size_t dset_idx = 0; dset_idx < num_dsets; dset_idx++) { /* Continually write random float data, followed by zeroed-out data */ if (i % 2) @@ -8544,6 +8603,12 @@ test_edge_chunks_no_overlap(const char *parent_group, H5Z_filter_t filter_id, hi filespace = H5Screate_simple(WRITE_UNSHARED_FILTERED_EDGE_CHUNKS_DATASET_DIMS, dataset_dims, NULL); VRFY((filespace >= 0), "File dataspace creation succeeded"); + /* + * Since we're only doing a partial write to the dataset, make + * sure the fill time is set appropriately + */ + VRFY((H5Pset_fill_time(plist_id, H5D_FILL_TIME_IFSET) >= 0), "H5Pset_fill_time succeeded"); + H5Pset_chunk_opts(plist_id, H5D_CHUNK_DONT_FILTER_PARTIAL_CHUNKS); /* Create datasets depending on the current test mode */ @@ -8748,6 +8813,12 @@ test_edge_chunks_overlap(const char *parent_group, H5Z_filter_t filter_id, hid_t filespace = H5Screate_simple(WRITE_SHARED_FILTERED_EDGE_CHUNKS_DATASET_DIMS, dataset_dims, NULL); VRFY((filespace >= 0), "File dataspace creation succeeded"); + /* + * Since we're only doing a partial write to the dataset, make + * sure the fill time is set appropriately + */ + VRFY((H5Pset_fill_time(plist_id, H5D_FILL_TIME_IFSET) >= 0), "H5Pset_fill_time succeeded"); + H5Pset_chunk_opts(plist_id, H5D_CHUNK_DONT_FILTER_PARTIAL_CHUNKS); /* Create datasets depending on the current test mode */ @@ -8870,6 +8941,9 @@ test_fill_values(const char *parent_group, H5Z_filter_t filter_id, hid_t fapl_id plist_id = H5Pcopy(dcpl_id); VRFY((plist_id >= 0), "DCPL copy succeeded"); + /* Make sure the fill time is set appropriately */ + VRFY((H5Pset_fill_time(plist_id, H5D_FILL_TIME_IFSET) >= 0), "H5Pset_fill_time succeeded"); + VRFY((H5Pset_chunk(plist_id, FILL_VALUES_TEST_DATASET_DIMS, chunk_dims) >= 0), "Chunk size set"); /* Add test filter to the pipeline */ @@ -9231,6 +9305,9 @@ test_fill_value_undefined(const char *parent_group, H5Z_filter_t filter_id, hid_ plist_id = H5Pcopy(dcpl_id); VRFY((plist_id >= 0), "DCPL copy succeeded"); + /* Make sure the fill time is set appropriately */ + VRFY((H5Pset_fill_time(plist_id, H5D_FILL_TIME_IFSET) >= 0), "H5Pset_fill_time succeeded"); + VRFY((H5Pset_chunk(plist_id, FILL_VALUE_UNDEFINED_TEST_DATASET_DIMS, chunk_dims) >= 0), "Chunk size set"); /* Add test filter to the pipeline */ @@ -9449,12 +9526,10 @@ test_fill_time_never(const char *parent_group, H5Z_filter_t filter_id, hid_t fap size_t num_dsets; hid_t dset_ids[MAX_NUM_DSETS_MULTI]; hid_t fspace_ids[MAX_NUM_DSETS_MULTI]; - hid_t file_id = H5I_INVALID_HID, plist_id = H5I_INVALID_HID; - hid_t group_id = H5I_INVALID_HID; - hid_t filespace = H5I_INVALID_HID; - int *recvcounts = NULL; - int *displs = NULL; - int mpi_code; + hid_t file_id = H5I_INVALID_HID; + hid_t plist_id = H5I_INVALID_HID; + hid_t group_id = H5I_INVALID_HID; + hid_t filespace = H5I_INVALID_HID; if (MAINPROCESS) puts("Testing fill time H5D_FILL_TIME_NEVER"); @@ -9504,7 +9579,7 @@ test_fill_time_never(const char *parent_group, H5Z_filter_t filter_id, hid_t fap VRFY((set_dcpl_filter(plist_id, filter_id, NULL) >= 0), "Filter set"); /* Set a fill value */ - fill_value = FILL_VALUES_TEST_FILL_VAL; + fill_value = FILL_TIME_NEVER_TEST_FILL_VAL; VRFY((H5Pset_fill_value(plist_id, HDF5_DATATYPE_NAME, &fill_value) >= 0), "Fill Value set"); /* Set fill time of 'never' */ @@ -9519,6 +9594,21 @@ test_fill_time_never(const char *parent_group, H5Z_filter_t filter_id, hid_t fap VRFY((H5Sclose(filespace) >= 0), "File dataspace close succeeded"); + /* Allocate buffer for reading entire dataset */ + read_buf_size = dataset_dims[0] * dataset_dims[1] * sizeof(C_DATATYPE); + + for (size_t dset_idx = 0; dset_idx < num_dsets; dset_idx++) { + read_bufs[dset_idx] = calloc(1, read_buf_size); + VRFY((NULL != read_bufs[dset_idx]), "calloc succeeded"); + } + + /* Allocate buffer of fill values */ + fill_buf = calloc(1, read_buf_size); + VRFY((NULL != fill_buf), "calloc succeeded"); + + for (size_t i = 0; i < read_buf_size / sizeof(C_DATATYPE); i++) + fill_buf[i] = FILL_TIME_NEVER_TEST_FILL_VAL; + /* * Since we aren't writing fill values to the chunks of the * datasets we just created, close and re-open file to ensure @@ -9538,37 +9628,21 @@ test_fill_time_never(const char *parent_group, H5Z_filter_t filter_id, hid_t fap open_datasets(group_id, FILL_TIME_NEVER_TEST_DATASET_NAME, num_dsets, test_mode, dset_ids); - /* Allocate buffer for reading entire dataset */ - read_buf_size = dataset_dims[0] * dataset_dims[1] * sizeof(C_DATATYPE); - - for (size_t dset_idx = 0; dset_idx < num_dsets; dset_idx++) { - read_bufs[dset_idx] = calloc(1, read_buf_size); - VRFY((NULL != read_bufs[dset_idx]), "calloc succeeded"); - } - - fill_buf = calloc(1, read_buf_size); - VRFY((NULL != fill_buf), "calloc succeeded"); - - /* Read entire dataset and verify that the fill value isn't returned */ - read_datasets(num_dsets, dset_ids, HDF5_DATATYPE_NAME, H5S_ALL, H5S_ALL, dxpl_id, read_bufs, test_mode); - - for (size_t i = 0; i < read_buf_size / sizeof(C_DATATYPE); i++) - fill_buf[i] = FILL_TIME_NEVER_TEST_FILL_VAL; - /* - * It should be very unlikely for the dataset's random - * values to all be the fill value, so this should be - * a safe comparison in theory. + * Read entire dataset just to try to verify bad behavior doesn't + * occur. Don't attempt to verify the contents of the read buffer(s) + * yet, because there's no guarantee as to what may have been + * read from the dataset. */ - for (size_t dset_idx = 0; dset_idx < num_dsets; dset_idx++) - VRFY((0 != memcmp(read_bufs[dset_idx], fill_buf, read_buf_size)), "Data verification succeeded"); + read_datasets(num_dsets, dset_ids, HDF5_DATATYPE_NAME, H5S_ALL, H5S_ALL, dxpl_id, read_bufs, test_mode); /* * Write to part of the first chunk in the dataset with - * all ranks, then read the whole dataset and ensure that - * the fill value isn't returned for the unwritten part of - * the chunk, as well as for the rest of the dataset that - * hasn't been written to yet. + * all ranks, then read the whole dataset just to try to + * verify bad behavior doesn't occur. Don't attempt to + * verify the contents of the read buffer(s) yet, because + * there's no guarantee as to what may have been read from + * the dataset. */ count[0] = 1; count[1] = 1; @@ -9609,34 +9683,6 @@ test_fill_time_never(const char *parent_group, H5Z_filter_t filter_id, hid_t fap read_datasets(num_dsets, dset_ids, HDF5_DATATYPE_NAME, H5S_ALL, H5S_ALL, dxpl_id, read_bufs, test_mode); - /* - * Each MPI rank communicates their written piece of data - * into each other rank's correctness-checking buffer - */ - recvcounts = calloc(1, (size_t)mpi_size * sizeof(*recvcounts)); - VRFY((NULL != recvcounts), "calloc succeeded"); - - displs = calloc(1, (size_t)mpi_size * sizeof(*displs)); - VRFY((NULL != displs), "calloc succeeded"); - - for (size_t i = 0; i < (size_t)mpi_size; i++) { - recvcounts[i] = (int)(count[1] * block[1]); - displs[i] = (int)(i * dataset_dims[1]); - } - - for (size_t dset_idx = 0; dset_idx < num_dsets; dset_idx++) { - mpi_code = MPI_Allgatherv(data_bufs[dset_idx], recvcounts[mpi_rank], C_DATATYPE_MPI, fill_buf, - recvcounts, displs, C_DATATYPE_MPI, comm); - VRFY((MPI_SUCCESS == mpi_code), "MPI_Allgatherv succeeded"); - - /* - * It should be very unlikely for the dataset's random - * values to all be the fill value, so this should be - * a safe comparison in theory. - */ - VRFY((0 != memcmp(read_bufs[dset_idx], fill_buf, read_buf_size)), "Data verification succeeded"); - } - for (size_t dset_idx = 0; dset_idx < num_dsets; dset_idx++) VRFY((H5Sclose(fspace_ids[dset_idx]) >= 0), "File dataspace close succeeded"); @@ -9680,9 +9726,6 @@ test_fill_time_never(const char *parent_group, H5Z_filter_t filter_id, hid_t fap VRFY((tmp_buf[j] != FILL_TIME_NEVER_TEST_FILL_VAL), "Data verification succeeded"); } - free(displs); - free(recvcounts); - free(fill_buf); for (size_t dset_idx = 0; dset_idx < num_dsets; dset_idx++) { @@ -9827,7 +9870,13 @@ main(int argc, char **argv) fcpl_id = H5Pcreate(H5P_FILE_CREATE); VRFY((fcpl_id >= 0), "FCPL creation succeeded"); - VRFY((H5Pset_file_space_strategy(fcpl_id, H5F_FSPACE_STRATEGY_PAGE, true, 1) >= 0), + /* + * TODO: Ideally, use persistent free space management. However, + * this occasionally runs into an infinite loop in the library's + * free space management code, so don't persist free space for now + * until that is fixed. + */ + VRFY((H5Pset_file_space_strategy(fcpl_id, H5F_FSPACE_STRATEGY_PAGE, false, 1) >= 0), "H5Pset_file_space_strategy succeeded"); VRFY((h5_fixname(FILENAME[0], fapl_id, filenames[0], sizeof(filenames[0])) != NULL), @@ -9956,6 +10005,19 @@ main(int argc, char **argv) continue; } + /* + * If TestExpress is > 1, only run the multi-chunk I/O + * configuration tests for the 'USE_SINGLE_DATASET' case, + * as the 'USE_MULTIPLE_DATASETS' and 'USE_MULTIPLE_DATASETS_MIXED_FILTERED' + * cases are more stressful on the file system. + */ + if (test_express_level_g > 1) { + if (((test_mode == USE_MULTIPLE_DATASETS) || + (test_mode == USE_MULTIPLE_DATASETS_MIXED_FILTERED)) && + (chunk_opt != H5FD_MPIO_CHUNK_ONE_IO)) + continue; + } + if (MAINPROCESS) { printf("== Running tests in mode '%s' with filter '%s' using selection I/O mode " "'%s', '%s' and '%s' allocation time ==\n\n", @@ -9992,6 +10054,15 @@ main(int argc, char **argv) VRFY((H5Pset_dxpl_mpio_chunk_opt(dxpl_id, chunk_opt) >= 0), "H5Pset_dxpl_mpio_chunk_opt succeeded"); + /* + * Disable writing of fill values by default. Otherwise, a + * lot of time may be spent writing fill values to chunks + * when they're going to be fully overwritten anyway. + * Individual tests will alter this behavior as necessary. + */ + VRFY((H5Pset_fill_time(dcpl_id, H5D_FILL_TIME_NEVER) >= 0), + "H5Pset_fill_time succeeded"); + /* Create a group to hold all the datasets for this combination * of filter and chunk optimization mode. Then, close the file * again since some tests may need to open the file in a special diff --git a/testpar/t_filters_parallel.h b/testpar/t_filters_parallel.h index c0b1db8..04d3639 100644 --- a/testpar/t_filters_parallel.h +++ b/testpar/t_filters_parallel.h @@ -444,7 +444,7 @@ typedef struct { #define SHRINKING_GROWING_CHUNKS_NCOLS (mpi_size * DIM1_SCALE_FACTOR) #define SHRINKING_GROWING_CHUNKS_CH_NROWS (SHRINKING_GROWING_CHUNKS_NROWS / mpi_size) #define SHRINKING_GROWING_CHUNKS_CH_NCOLS (SHRINKING_GROWING_CHUNKS_NCOLS / mpi_size) -#define SHRINKING_GROWING_CHUNKS_NLOOPS 20 +#define SHRINKING_GROWING_CHUNKS_NLOOPS 8 /* Defines for the unshared filtered edge chunks write test */ #define WRITE_UNSHARED_FILTERED_EDGE_CHUNKS_DATASET_NAME "unshared_filtered_edge_chunks_write" |