diff options
author | Larry Knox <lrknox@hdfgroup.org> | 2021-09-14 13:37:38 (GMT) |
---|---|---|
committer | GitHub <noreply@github.com> | 2021-09-14 13:37:38 (GMT) |
commit | d110a16912657d45fcc284305daa2cd69322c5ba (patch) | |
tree | 0389ef34390b84805f919ef7ca841e1555221879 /testpar | |
parent | b30a44973e97c56bf7772cdbf767b98f84122397 (diff) | |
download | hdf5-d110a16912657d45fcc284305daa2cd69322c5ba.zip hdf5-d110a16912657d45fcc284305daa2cd69322c5ba.tar.gz hdf5-d110a16912657d45fcc284305daa2cd69322c5ba.tar.bz2 |
H5Ztrans: (feature) Improved H5Z_xform_noop() and H5Z_xform_create() … (#933) (#1008)
* H5Ztrans: (feature) Improved H5Z_xform_noop() and H5Z_xform_create() function
- Made a small improvement for H5Z_xform_noop() function. Now,
the function will also return TRUE if the data transform function
expression = "x". For this case, the HDF5 library behaves in a
similar fashion as the case when no data transform function has been
specified.
- Improved the inline documentation of the function
H5Z_xform_create() such it is more inline with the rest of the
code.
* Committing clang-format changes
* H5Ztrans: (feature) Added 3 tests for improved H5Z_xform_noop() function
- Added serial test with data transform expression = "x" to
verify the improved H5Z_xform_noop() function behaves as expected.
- Added 2 parallel tests with data transform expression = "x"
in combination with a filter. Before, these tests will fail but
with the improved H5Z_xform_noop() function they work and result
in the expected behavior.
- Small bug fix for one of parallel filter tests.
* Committing clang-format changes
* H5Ztrans: (feature) Added release note about detection of the
simple data transform function "x".
- Added a brief explanation about the implemented improvement
of the detection of the simple data transform function "x"
to the RELEASE.txt file.
Co-authored-by: github-actions <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Jan-Willem Blokland <Jan-Willem.Blokland@Shell.com>
Co-authored-by: github-actions <41898282+github-actions[bot]@users.noreply.github.com>
Diffstat (limited to 'testpar')
-rw-r--r-- | testpar/t_filters_parallel.c | 403 | ||||
-rw-r--r-- | testpar/t_filters_parallel.h | 20 |
2 files changed, 422 insertions, 1 deletions
diff --git a/testpar/t_filters_parallel.c b/testpar/t_filters_parallel.c index 50cd306..ffeddbe 100644 --- a/testpar/t_filters_parallel.c +++ b/testpar/t_filters_parallel.c @@ -62,6 +62,7 @@ static void test_write_filtered_dataset_single_no_selection(void); static void test_write_filtered_dataset_all_no_selection(void); static void test_write_filtered_dataset_point_selection(void); static void test_write_filtered_dataset_interleaved_write(void); +static void test_write_transformed_filtered_dataset_no_overlap(void); static void test_write_3d_filtered_dataset_no_overlap_separate_pages(void); static void test_write_3d_filtered_dataset_no_overlap_same_pages(void); static void test_write_3d_filtered_dataset_overlap(void); @@ -79,6 +80,7 @@ static void test_read_filtered_dataset_single_no_selection(void); static void test_read_filtered_dataset_all_no_selection(void); static void test_read_filtered_dataset_point_selection(void); static void test_read_filtered_dataset_interleaved_read(void); +static void test_read_transformed_filtered_dataset_no_overlap(void); static void test_read_3d_filtered_dataset_no_overlap_separate_pages(void); static void test_read_3d_filtered_dataset_no_overlap_same_pages(void); static void test_read_3d_filtered_dataset_overlap(void); @@ -120,6 +122,7 @@ static void (*tests[])(void) = { test_write_filtered_dataset_all_no_selection, test_write_filtered_dataset_point_selection, test_write_filtered_dataset_interleaved_write, + test_write_transformed_filtered_dataset_no_overlap, test_write_3d_filtered_dataset_no_overlap_separate_pages, test_write_3d_filtered_dataset_no_overlap_same_pages, test_write_3d_filtered_dataset_overlap, @@ -135,6 +138,7 @@ static void (*tests[])(void) = { test_read_filtered_dataset_all_no_selection, test_read_filtered_dataset_point_selection, test_read_filtered_dataset_interleaved_read, + test_read_transformed_filtered_dataset_no_overlap, test_read_3d_filtered_dataset_no_overlap_separate_pages, test_read_3d_filtered_dataset_no_overlap_same_pages, test_read_3d_filtered_dataset_overlap, @@ -428,7 +432,7 @@ test_write_filtered_dataset_no_overlap(void) /* Select hyperslab in the file */ filespace = H5Dget_space(dset_id); - VRFY((dset_id >= 0), "File dataspace retrieval succeeded"); + VRFY((filespace >= 0), "File dataspace retrieval succeeded"); VRFY((H5Sselect_hyperslab(filespace, H5S_SELECT_SET, start, stride, count, block) >= 0), "Hyperslab selection succeeded"); @@ -1288,6 +1292,177 @@ test_write_filtered_dataset_interleaved_write(void) } /* + * Tests parallel write of transformed and filtered data + * in the case where only one process is writing to a + * particular chunk in the operation. Normally, a data + * transform function will cause the parallel library to + * break to independent I/O and this isn't allowed when + * there are filters in the pipeline. However, in this + * case the parallel library recognizes that the used + * data transform function "x" is the same as not applying + * the transform function. Therefore it does not apply + * the transform function resulting in not breaking to + * independent I/O. + * + * Programmer: Jan-Willem Blokland + * 08/20/2021 + */ +static void +test_write_transformed_filtered_dataset_no_overlap(void) +{ + C_DATATYPE *data = NULL; + C_DATATYPE *read_buf = NULL; + C_DATATYPE *correct_buf = NULL; + hsize_t dataset_dims[WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS]; + hsize_t chunk_dims[WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS]; + hsize_t sel_dims[WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS]; + hsize_t start[WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS]; + hsize_t stride[WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS]; + hsize_t count[WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS]; + hsize_t block[WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS]; + size_t i, data_size, correct_buf_size; + hid_t file_id = -1, dset_id = -1, plist_id = -1; + hid_t filespace = -1, memspace = -1; + + if (MAINPROCESS) + HDputs("Testing write to unshared transformed and filtered chunks"); + + CHECK_CUR_FILTER_AVAIL(); + + /* Set up file access property list with parallel I/O access */ + plist_id = H5Pcreate(H5P_FILE_ACCESS); + VRFY((plist_id >= 0), "FAPL creation succeeded"); + + VRFY((H5Pset_fapl_mpio(plist_id, comm, info) >= 0), "Set FAPL MPIO succeeded"); + + VRFY((H5Pset_libver_bounds(plist_id, H5F_LIBVER_LATEST, H5F_LIBVER_LATEST) >= 0), + "Set libver bounds succeeded"); + + file_id = H5Fopen(filenames[0], H5F_ACC_RDWR, plist_id); + VRFY((file_id >= 0), "Test file open succeeded"); + + VRFY((H5Pclose(plist_id) >= 0), "FAPL close succeeded"); + + /* Create the dataspace for the dataset */ + dataset_dims[0] = (hsize_t)WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NROWS; + dataset_dims[1] = (hsize_t)WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NCOLS; + chunk_dims[0] = (hsize_t)WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NROWS; + chunk_dims[1] = (hsize_t)WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NCOLS; + sel_dims[0] = (hsize_t)WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NROWS; + sel_dims[1] = (hsize_t)WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NCOLS; + + filespace = H5Screate_simple(WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS, dataset_dims, NULL); + VRFY((filespace >= 0), "File dataspace creation succeeded"); + + memspace = H5Screate_simple(WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS, sel_dims, NULL); + VRFY((memspace >= 0), "Memory dataspace creation succeeded"); + + /* Create chunked dataset */ + plist_id = H5Pcreate(H5P_DATASET_CREATE); + VRFY((plist_id >= 0), "DCPL creation succeeded"); + + VRFY((H5Pset_chunk(plist_id, WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS, chunk_dims) >= 0), + "Chunk size set"); + + /* Add test filter to the pipeline */ + VRFY((set_dcpl_filter(plist_id) >= 0), "Filter set"); + + dset_id = H5Dcreate2(file_id, WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_NAME, HDF5_DATATYPE_NAME, + filespace, H5P_DEFAULT, plist_id, H5P_DEFAULT); + VRFY((dset_id >= 0), "Dataset creation succeeded"); + + VRFY((H5Pclose(plist_id) >= 0), "DCPL close succeeded"); + VRFY((H5Sclose(filespace) >= 0), "File dataspace close succeeded"); + + /* Each process defines the dataset selection in memory and writes + * it to the hyperslab in the file + */ + count[0] = 1; + count[1] = (hsize_t)WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NCOLS / + (hsize_t)WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NCOLS; + stride[0] = (hsize_t)WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NROWS; + stride[1] = (hsize_t)WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NCOLS; + block[0] = (hsize_t)WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NROWS; + block[1] = (hsize_t)WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NCOLS; + start[0] = ((hsize_t)mpi_rank * (hsize_t)WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NROWS * count[0]); + start[1] = 0; + + if (VERBOSE_MED) { + HDprintf("Process %d is writing with count[ %" PRIuHSIZE ", %" PRIuHSIZE " ], stride[ %" PRIuHSIZE + ", %" PRIuHSIZE " ], start[ %" PRIuHSIZE ", %" PRIuHSIZE " ], block size[ %" PRIuHSIZE + ", %" PRIuHSIZE " ]\n", + mpi_rank, count[0], count[1], stride[0], stride[1], start[0], start[1], block[0], block[1]); + HDfflush(stdout); + } + + /* Select hyperslab in the file */ + filespace = H5Dget_space(dset_id); + VRFY((filespace >= 0), "File dataspace retrieval succeeded"); + + VRFY((H5Sselect_hyperslab(filespace, H5S_SELECT_SET, start, stride, count, block) >= 0), + "Hyperslab selection succeeded"); + + /* Fill data buffer */ + data_size = sel_dims[0] * sel_dims[1] * sizeof(*data); + correct_buf_size = dataset_dims[0] * dataset_dims[1] * sizeof(*correct_buf); + + data = (C_DATATYPE *)HDcalloc(1, data_size); + VRFY((NULL != data), "HDcalloc succeeded"); + + correct_buf = (C_DATATYPE *)HDcalloc(1, correct_buf_size); + VRFY((NULL != correct_buf), "HDcalloc succeeded"); + + for (i = 0; i < data_size / sizeof(*data); i++) + data[i] = (C_DATATYPE)GEN_DATA(i); + + for (i = 0; i < correct_buf_size / sizeof(*correct_buf); i++) + correct_buf[i] = (C_DATATYPE)((i % (dataset_dims[0] / (hsize_t)mpi_size * dataset_dims[1])) + + (i / (dataset_dims[0] / (hsize_t)mpi_size * dataset_dims[1]))); + + /* Create property list for collective dataset write and data transform */ + plist_id = H5Pcreate(H5P_DATASET_XFER); + VRFY((plist_id >= 0), "DXPL creation succeeded"); + + VRFY((H5Pset_dxpl_mpio(plist_id, H5FD_MPIO_COLLECTIVE) >= 0), "Set DXPL MPIO succeeded"); + + /* Set data transform expression */ + VRFY((H5Pset_data_transform(plist_id, "x") >= 0), "Set data transform expression succeeded"); + + VRFY((H5Dwrite(dset_id, HDF5_DATATYPE_NAME, memspace, filespace, plist_id, data) >= 0), + "Dataset write succeeded"); + + if (data) + HDfree(data); + + VRFY((H5Dclose(dset_id) >= 0), "Dataset close succeeded"); + + /* Verify the correct data was written */ + read_buf = (C_DATATYPE *)HDcalloc(1, correct_buf_size); + VRFY((NULL != read_buf), "HDcalloc succeeded"); + + dset_id = H5Dopen2(file_id, "/" WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_NAME, H5P_DEFAULT); + VRFY((dset_id >= 0), "Dataset open succeeded"); + + VRFY((H5Dread(dset_id, HDF5_DATATYPE_NAME, H5S_ALL, H5S_ALL, plist_id, read_buf) >= 0), + "Dataset read succeeded"); + + VRFY((0 == HDmemcmp(read_buf, correct_buf, correct_buf_size)), "Data verification succeeded"); + + if (correct_buf) + HDfree(correct_buf); + if (read_buf) + HDfree(read_buf); + + VRFY((H5Dclose(dset_id) >= 0), "Dataset close succeeded"); + VRFY((H5Sclose(filespace) >= 0), "File dataspace close succeeded"); + VRFY((H5Sclose(memspace) >= 0), "Memory dataspace close succeeded"); + VRFY((H5Pclose(plist_id) >= 0), "DXPL close succeeded"); + VRFY((H5Fclose(file_id) >= 0), "File close succeeded"); + + return; +} + +/* * Tests parallel write of filtered data in the case where * the dataset has 3 dimensions and each process writes * to its own "page" in the 3rd dimension. @@ -4224,6 +4399,232 @@ test_read_3d_filtered_dataset_no_overlap_separate_pages(void) } /* + * Tests parallel read of transformed and filtered data in the + * case where only one process is reading from a particular + * chunk in the operation. Normally, a data transform function + * will cause the parallel library to break to independent I/O + * and this isn't allowed when there are filters in the pipeline. + * However, in this case the parallel library recognizes that + * the used data transform function "x" is the same as not + * applying the transform function. Therefore it does not apply + * the transform function resulting in not breaking to + * independent I/O. + * + * The MAINPROCESS rank will first write out all of the + * data to the dataset. Then, each rank reads a part of + * the dataset and contributes its piece to a global buffer + * that is checked for consistency. + * + * Programmer: Jan-Willem Blokland + * 08/20/2021 + */ +static void +test_read_transformed_filtered_dataset_no_overlap(void) +{ + C_DATATYPE *read_buf = NULL; + C_DATATYPE *correct_buf = NULL; + C_DATATYPE *global_buf = NULL; + hsize_t dataset_dims[READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS]; + hsize_t chunk_dims[READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS]; + hsize_t sel_dims[READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS]; + hsize_t start[READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS]; + hsize_t stride[READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS]; + hsize_t count[READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS]; + hsize_t block[READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS]; + hsize_t flat_dims[1]; + size_t i, read_buf_size, correct_buf_size; + hid_t file_id = -1, dset_id = -1, plist_id = -1; + hid_t filespace = -1, memspace = -1; + int * recvcounts = NULL; + int * displs = NULL; + + if (MAINPROCESS) + HDputs("Testing read from unshared transformed and filtered chunks"); + + CHECK_CUR_FILTER_AVAIL(); + + dataset_dims[0] = (hsize_t)READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NROWS; + dataset_dims[1] = (hsize_t)READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NCOLS; + + /* Setup the buffer for writing and for comparison */ + correct_buf_size = (hsize_t)READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NROWS * + (hsize_t)READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NCOLS * sizeof(*correct_buf); + + correct_buf = (C_DATATYPE *)HDcalloc(1, correct_buf_size); + VRFY((NULL != correct_buf), "HDcalloc succeeded"); + + for (i = 0; i < correct_buf_size / sizeof(*correct_buf); i++) + correct_buf[i] = (C_DATATYPE)((i % (dataset_dims[0] / (hsize_t)mpi_size * dataset_dims[1])) + + (i / (dataset_dims[0] / (hsize_t)mpi_size * dataset_dims[1]))); + + if (MAINPROCESS) { + plist_id = H5Pcreate(H5P_FILE_ACCESS); + VRFY((plist_id >= 0), "FAPL creation succeeded"); + + VRFY((H5Pset_libver_bounds(plist_id, H5F_LIBVER_LATEST, H5F_LIBVER_LATEST) >= 0), + "Set libver bounds succeeded"); + + file_id = H5Fopen(filenames[0], H5F_ACC_RDWR, plist_id); + VRFY((file_id >= 0), "Test file open succeeded"); + + VRFY((H5Pclose(plist_id) >= 0), "FAPL close succeeded"); + + /* Create the dataspace for the dataset */ + filespace = + H5Screate_simple(READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS, dataset_dims, NULL); + VRFY((filespace >= 0), "File dataspace creation succeeded"); + + /* Create chunked dataset */ + chunk_dims[0] = (hsize_t)READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NROWS; + chunk_dims[1] = (hsize_t)READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NCOLS; + + plist_id = H5Pcreate(H5P_DATASET_CREATE); + VRFY((plist_id >= 0), "DCPL creation succeeded"); + + VRFY( + (H5Pset_chunk(plist_id, READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS, chunk_dims) >= 0), + "Chunk size set"); + + /* Add test filter to the pipeline */ + VRFY((set_dcpl_filter(plist_id) >= 0), "Filter set"); + + dset_id = H5Dcreate2(file_id, READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_NAME, + HDF5_DATATYPE_NAME, filespace, H5P_DEFAULT, plist_id, H5P_DEFAULT); + VRFY((dset_id >= 0), "Dataset creation succeeded"); + + VRFY((H5Pclose(plist_id) >= 0), "DCPL close succeeded"); + VRFY((H5Sclose(filespace) >= 0), "File dataspace close succeeded"); + + /* Create property list for collective dataset read */ + plist_id = H5Pcreate(H5P_DATASET_XFER); + VRFY((plist_id >= 0), "DXPL creation succeeded"); + + /* Set data transform expression */ + VRFY((H5Pset_data_transform(plist_id, "x") >= 0), "Set data transform expression succeeded"); + + VRFY((H5Dwrite(dset_id, HDF5_DATATYPE_NAME, H5S_ALL, H5S_ALL, plist_id, correct_buf) >= 0), + "Dataset write succeeded"); + + VRFY((H5Pclose(plist_id) >= 0), "DXPL close succeeded"); + VRFY((H5Dclose(dset_id) >= 0), "Dataset close succeeded"); + VRFY((H5Fclose(file_id) >= 0), "File close succeeded"); + } + + /* Set up file access property list with parallel I/O access */ + plist_id = H5Pcreate(H5P_FILE_ACCESS); + VRFY((plist_id >= 0), "FAPL creation succeeded"); + + VRFY((H5Pset_fapl_mpio(plist_id, comm, info) >= 0), "Set FAPL MPIO succeeded"); + + VRFY((H5Pset_libver_bounds(plist_id, H5F_LIBVER_LATEST, H5F_LIBVER_LATEST) >= 0), + "Set libver bounds succeeded"); + + file_id = H5Fopen(filenames[0], H5F_ACC_RDONLY, plist_id); + VRFY((file_id >= 0), "Test file open succeeded"); + + VRFY((H5Pclose(plist_id) >= 0), "FAPL close succeeded"); + + dset_id = H5Dopen2(file_id, "/" READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_NAME, H5P_DEFAULT); + VRFY((dset_id >= 0), "Dataset open succeeded"); + + sel_dims[0] = (hsize_t)READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NROWS; + sel_dims[1] = (hsize_t)READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NCOLS; + + /* Setup one-dimensional memory dataspace for reading the dataset data into a contiguous buffer */ + flat_dims[0] = sel_dims[0] * sel_dims[1]; + + memspace = H5Screate_simple(1, flat_dims, NULL); + VRFY((memspace >= 0), "Memory dataspace creation succeeded"); + + /* Select hyperslab in the file */ + filespace = H5Dget_space(dset_id); + VRFY((filespace >= 0), "File dataspace retrieval succeeded"); + + /* + * Each process defines the dataset selection in the file and reads + * it to the selection in memory + */ + count[0] = 1; + count[1] = (hsize_t)READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NCOLS / + (hsize_t)READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NCOLS; + stride[0] = (hsize_t)READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NROWS; + stride[1] = (hsize_t)READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NCOLS; + block[0] = (hsize_t)READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NROWS; + block[1] = (hsize_t)READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NCOLS; + start[0] = ((hsize_t)mpi_rank * (hsize_t)READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NROWS * count[0]); + start[1] = 0; + + if (VERBOSE_MED) { + HDprintf("Process %d is reading with count[ %" PRIuHSIZE ", %" PRIuHSIZE " ], stride[ %" PRIuHSIZE + ", %" PRIuHSIZE " ], start[ %" PRIuHSIZE ", %" PRIuHSIZE " ], block size[ %" PRIuHSIZE + ", %" PRIuHSIZE " ]\n", + mpi_rank, count[0], count[1], stride[0], stride[1], start[0], start[1], block[0], block[1]); + HDfflush(stdout); + } + + VRFY((H5Sselect_hyperslab(filespace, H5S_SELECT_SET, start, stride, count, block) >= 0), + "Hyperslab selection succeeded"); + + /* Create property list for collective dataset read and data transform */ + plist_id = H5Pcreate(H5P_DATASET_XFER); + VRFY((plist_id >= 0), "DXPL creation succeeded"); + + VRFY((H5Pset_dxpl_mpio(plist_id, H5FD_MPIO_COLLECTIVE) >= 0), "Set DXPL MPIO succeeded"); + + /* Set data transform expression */ + VRFY((H5Pset_data_transform(plist_id, "x") >= 0), "Set data transform expression succeeded"); + + read_buf_size = flat_dims[0] * sizeof(*read_buf); + + read_buf = (C_DATATYPE *)HDcalloc(1, read_buf_size); + VRFY((NULL != read_buf), "HDcalloc succeeded"); + + VRFY((H5Dread(dset_id, HDF5_DATATYPE_NAME, memspace, filespace, plist_id, read_buf) >= 0), + "Dataset read succeeded"); + + global_buf = (C_DATATYPE *)HDcalloc(1, correct_buf_size); + VRFY((NULL != global_buf), "HDcalloc succeeded"); + + /* Collect each piece of data from all ranks into a global buffer on all ranks */ + recvcounts = (int *)HDcalloc(1, (size_t)mpi_size * sizeof(*recvcounts)); + VRFY((NULL != recvcounts), "HDcalloc succeeded"); + + for (i = 0; i < (size_t)mpi_size; i++) + recvcounts[i] = (int)flat_dims[0]; + + displs = (int *)HDcalloc(1, (size_t)mpi_size * sizeof(*displs)); + VRFY((NULL != displs), "HDcalloc succeeded"); + + for (i = 0; i < (size_t)mpi_size; i++) + displs[i] = (int)(i * flat_dims[0]); + + VRFY((MPI_SUCCESS == MPI_Allgatherv(read_buf, (int)flat_dims[0], C_DATATYPE_MPI, global_buf, recvcounts, + displs, C_DATATYPE_MPI, comm)), + "MPI_Allgatherv succeeded"); + + VRFY((0 == HDmemcmp(global_buf, correct_buf, correct_buf_size)), "Data verification succeeded"); + + if (displs) + HDfree(displs); + if (recvcounts) + HDfree(recvcounts); + if (global_buf) + HDfree(global_buf); + if (read_buf) + HDfree(read_buf); + if (correct_buf) + HDfree(correct_buf); + + VRFY((H5Dclose(dset_id) >= 0), "Dataset close succeeded"); + VRFY((H5Sclose(filespace) >= 0), "File dataspace close succeeded"); + VRFY((H5Sclose(memspace) >= 0), "Memory dataspace close succeeded"); + VRFY((H5Pclose(plist_id) >= 0), "DXPL close succeeded"); + VRFY((H5Fclose(file_id) >= 0), "File close succeeded"); + + return; +} + +/* * Tests parallel read of filtered data in the case where * the dataset has 3 dimensions and each process reads from * each "page" in the 3rd dimension. However, no chunk on a diff --git a/testpar/t_filters_parallel.h b/testpar/t_filters_parallel.h index 3804c09..7eb34ed 100644 --- a/testpar/t_filters_parallel.h +++ b/testpar/t_filters_parallel.h @@ -138,6 +138,16 @@ typedef struct { #define INTERLEAVED_WRITE_FILTERED_DATASET_NCOLS \ (INTERLEAVED_WRITE_FILTERED_DATASET_CH_NCOLS * DIM1_SCALE_FACTOR) +/* Defines for the unshared transformed and filtered chunks write test */ +#define WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_NAME "unshared_transformed_filtered_chunks_write" +#define WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS 2 +#define WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NROWS (mpi_size * DIM0_SCALE_FACTOR) +#define WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NCOLS (mpi_size * DIM1_SCALE_FACTOR) +#define WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NROWS \ + (WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NROWS / mpi_size) +#define WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NCOLS \ + (WRITE_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NCOLS / mpi_size) + /* Defines for the 3D unshared filtered dataset separate page write test */ #define WRITE_UNSHARED_FILTERED_CHUNKS_3D_SEP_PAGE_DATASET_NAME \ "3D_unshared_filtered_chunks_separate_pages_write" @@ -280,6 +290,16 @@ typedef struct { #define INTERLEAVED_READ_FILTERED_DATASET_NCOLS \ (INTERLEAVED_READ_FILTERED_DATASET_CH_NCOLS * DIM1_SCALE_FACTOR) +/* Defines for the unshared transformed and filtered chunks read test */ +#define READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_NAME "unshared_transformed_filtered_chunks_read" +#define READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_DATASET_DIMS 2 +#define READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NROWS (mpi_size * DIM0_SCALE_FACTOR) +#define READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NCOLS (mpi_size * DIM1_SCALE_FACTOR) +#define READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NROWS \ + (READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NROWS / mpi_size) +#define READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_CH_NCOLS \ + (READ_UNSHARED_TRANSFORMED_FILTERED_CHUNKS_NCOLS / mpi_size) + /* Defines for the 3D unshared filtered dataset separate page read test */ #define READ_UNSHARED_FILTERED_CHUNKS_3D_SEP_PAGE_DATASET_NAME \ "3D_unshared_filtered_chunks_separate_pages_read" |