diff options
Diffstat (limited to 'src/H5Dmpio.c')
| -rw-r--r-- | src/H5Dmpio.c | 78 |
1 files changed, 41 insertions, 37 deletions
diff --git a/src/H5Dmpio.c b/src/H5Dmpio.c index f5da33d..01cf932 100644 --- a/src/H5Dmpio.c +++ b/src/H5Dmpio.c @@ -800,6 +800,10 @@ H5D__chunk_collective_io(H5D_io_info_t *io_info, const H5D_type_info_t *type_inf HDassert(type_info); HDassert(fm); + /* Disable collective metadata reads for chunked dataset I/O operations + * in order to prevent potential hangs */ + H5CX_set_coll_metadata_read(FALSE); + /* Check the optional property list for the collective chunk IO optimization option */ if(H5CX_get_mpio_chunk_opt_mode(&chunk_opt_mode) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTGET, FAIL, "couldn't get chunk optimization option") @@ -2313,7 +2317,7 @@ if(H5DEBUG(D)) /* Broadcasting the MPI_IO option info. and chunk address info. */ if(MPI_SUCCESS != (mpi_code = MPI_Bcast(total_chunk_addr_array, (int)(sizeof(haddr_t) * fm->layout->u.chunk.nchunks), MPI_BYTE, (int)0, io_info->comm))) - HMPI_GOTO_ERROR(FAIL, "MPI_BCast failed", mpi_code) + HMPI_GOTO_ERROR(FAIL, "MPI_BCast failed", mpi_code) } /* end if */ /* Start at first node in chunk skip list */ @@ -2837,30 +2841,30 @@ H5D__chunk_redistribute_shared_chunks(const H5D_io_info_t *io_info, const H5D_ty size_t mod_data_size; /* Look up the chunk and get its file and memory dataspaces */ - if (NULL == (chunk_info = (H5D_chunk_info_t *) H5SL_search(fm->sel_chunks, &chunk_entry->index))) + if(NULL == (chunk_info = (H5D_chunk_info_t *) H5SL_search(fm->sel_chunks, &chunk_entry->index))) HGOTO_ERROR(H5E_DATASPACE, H5E_NOTFOUND, FAIL, "can't locate chunk in skip list") /* Determine size of serialized chunk file dataspace, plus the size of * the data being written */ - if (H5S_encode(chunk_info->fspace, &mod_data_p, &mod_data_size) < 0) + if(H5S_encode(chunk_info->fspace, &mod_data_p, &mod_data_size) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTENCODE, FAIL, "unable to get encoded dataspace size") - if ((iter_nelmts = H5S_GET_SELECT_NPOINTS(chunk_info->mspace)) < 0) + if((iter_nelmts = H5S_GET_SELECT_NPOINTS(chunk_info->mspace)) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTCOUNT, FAIL, "dataspace is invalid") mod_data_size += (size_t) iter_nelmts * type_info->src_type_size; - if (NULL == (mod_data[num_send_requests] = (unsigned char *) H5MM_malloc(mod_data_size))) + if(NULL == (mod_data[num_send_requests] = (unsigned char *) H5MM_malloc(mod_data_size))) HGOTO_ERROR(H5E_DATASET, H5E_CANTALLOC, FAIL, "couldn't allocate chunk modification send buffer") /* Serialize the chunk's file dataspace into the buffer */ mod_data_p = mod_data[num_send_requests]; - if (H5S_encode(chunk_info->fspace, &mod_data_p, &mod_data_size) < 0) + if(H5S_encode(chunk_info->fspace, &mod_data_p, &mod_data_size) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTENCODE, FAIL, "unable to encode dataspace") /* Initialize iterator for memory selection */ - if (H5S_select_iter_init(mem_iter, chunk_info->mspace, type_info->src_type_size) < 0) + if(H5S_select_iter_init(mem_iter, chunk_info->mspace, type_info->src_type_size) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTINIT, FAIL, "unable to initialize memory selection information") mem_iter_init = TRUE; @@ -2871,11 +2875,11 @@ H5D__chunk_redistribute_shared_chunks(const H5D_io_info_t *io_info, const H5D_ty /* Send modification data to new owner */ H5_CHECK_OVERFLOW(mod_data_size, size_t, int) H5_CHECK_OVERFLOW(chunk_entry->index, hsize_t, int) - if (MPI_SUCCESS != (mpi_code = MPI_Isend(mod_data[num_send_requests], (int) mod_data_size, MPI_BYTE, + if(MPI_SUCCESS != (mpi_code = MPI_Isend(mod_data[num_send_requests], (int) mod_data_size, MPI_BYTE, chunk_entry->owners.new_owner, (int) chunk_entry->index, io_info->comm, &send_requests[num_send_requests]))) HMPI_GOTO_ERROR(FAIL, "MPI_Isend failed", mpi_code) - if (mem_iter_init && H5S_SELECT_ITER_RELEASE(mem_iter) < 0) + if(mem_iter_init && H5S_SELECT_ITER_RELEASE(mem_iter) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTFREE, FAIL, "couldn't release memory selection iterator") mem_iter_init = FALSE; @@ -3044,11 +3048,11 @@ H5D__mpio_filtered_collective_write_type(H5D_filtered_collective_io_info_t *chun } /* end if */ done: - if (write_buf_array) + if(write_buf_array) H5MM_free(write_buf_array); - if (file_offset_array) + if(file_offset_array) H5MM_free(file_offset_array); - if (length_array) + if(length_array) H5MM_free(length_array); FUNC_LEAVE_NOAPI(ret_value) @@ -3166,7 +3170,7 @@ H5D__filtered_collective_chunk_entry_io(H5D_filtered_collective_io_info_t *chunk if (NULL == (mem_iter = (H5S_sel_iter_t *) H5MM_malloc(sizeof(H5S_sel_iter_t)))) HGOTO_ERROR(H5E_DATASET, H5E_CANTALLOC, FAIL, "couldn't allocate memory iterator") - if (H5S_select_iter_init(mem_iter, chunk_info->mspace, type_info->src_type_size) < 0) + if(H5S_select_iter_init(mem_iter, chunk_info->mspace, type_info->src_type_size) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTINIT, FAIL, "unable to initialize memory selection information") mem_iter_init = TRUE; @@ -3176,53 +3180,53 @@ H5D__filtered_collective_chunk_entry_io(H5D_filtered_collective_io_info_t *chunk * from the current process, then apply any modifications from other processes. Finally, * filter the newly-updated chunk. */ - switch (io_info->op_type) { + switch(io_info->op_type) { case H5D_IO_OP_READ: - if (NULL == (file_iter = (H5S_sel_iter_t *) H5MM_malloc(sizeof(H5S_sel_iter_t)))) + if(NULL == (file_iter = (H5S_sel_iter_t *) H5MM_malloc(sizeof(H5S_sel_iter_t)))) HGOTO_ERROR(H5E_DATASET, H5E_CANTALLOC, FAIL, "couldn't allocate file iterator") - if (H5S_select_iter_init(file_iter, chunk_info->fspace, type_info->src_type_size) < 0) + if(H5S_select_iter_init(file_iter, chunk_info->fspace, type_info->src_type_size) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTINIT, FAIL, "unable to initialize memory selection information") file_iter_init = TRUE; - if ((iter_nelmts = H5S_GET_SELECT_NPOINTS(chunk_info->fspace)) < 0) + if((iter_nelmts = H5S_GET_SELECT_NPOINTS(chunk_info->fspace)) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTCOUNT, FAIL, "dataspace is invalid") - if (NULL == (tmp_gath_buf = H5MM_malloc((hsize_t) iter_nelmts * type_info->src_type_size))) + if(NULL == (tmp_gath_buf = H5MM_malloc((hsize_t) iter_nelmts * type_info->src_type_size))) HGOTO_ERROR(H5E_DATASET, H5E_CANTALLOC, FAIL, "couldn't allocate temporary gather buffer") - if (!H5D__gather_mem(chunk_entry->buf, chunk_info->fspace, file_iter, (size_t) iter_nelmts, tmp_gath_buf)) + if(!H5D__gather_mem(chunk_entry->buf, chunk_info->fspace, file_iter, (size_t) iter_nelmts, tmp_gath_buf)) HGOTO_ERROR(H5E_IO, H5E_READERROR, FAIL, "couldn't gather from chunk buffer") - if ((iter_nelmts = H5S_GET_SELECT_NPOINTS(chunk_info->mspace)) < 0) + if((iter_nelmts = H5S_GET_SELECT_NPOINTS(chunk_info->mspace)) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTCOUNT, FAIL, "dataspace is invalid") - if (H5D__scatter_mem(tmp_gath_buf, chunk_info->mspace, mem_iter, (size_t) iter_nelmts, io_info->u.rbuf) < 0) - HGOTO_ERROR(H5E_DATASET, H5E_READERROR, FAIL, "couldn't scatter to read buffer") + if(H5D__scatter_mem(tmp_gath_buf, chunk_info->mspace, mem_iter, (size_t) iter_nelmts, io_info->u.rbuf) < 0) + HGOTO_ERROR(H5E_DATASET, H5E_READERROR, FAIL, "couldn't scatter to read buffer") break; case H5D_IO_OP_WRITE: - if ((iter_nelmts = H5S_GET_SELECT_NPOINTS(chunk_info->mspace)) < 0) + if((iter_nelmts = H5S_GET_SELECT_NPOINTS(chunk_info->mspace)) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTCOUNT, FAIL, "dataspace is invalid") - if (NULL == (tmp_gath_buf = H5MM_malloc((hsize_t) iter_nelmts * type_info->src_type_size))) + if(NULL == (tmp_gath_buf = H5MM_malloc((hsize_t) iter_nelmts * type_info->src_type_size))) HGOTO_ERROR(H5E_DATASET, H5E_CANTALLOC, FAIL, "couldn't allocate temporary gather buffer") /* Gather modification data from the application write buffer into a temporary buffer */ if(!H5D__gather_mem(io_info->u.wbuf, chunk_info->mspace, mem_iter, (size_t) iter_nelmts, tmp_gath_buf)) HGOTO_ERROR(H5E_IO, H5E_WRITEERROR, FAIL, "couldn't gather from write buffer") - if (H5S_SELECT_ITER_RELEASE(mem_iter) < 0) + if(H5S_SELECT_ITER_RELEASE(mem_iter) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTFREE, FAIL, "couldn't release selection iterator") mem_iter_init = FALSE; /* Initialize iterator for file selection */ - if (H5S_select_iter_init(mem_iter, chunk_info->fspace, type_info->dst_type_size) < 0) + if(H5S_select_iter_init(mem_iter, chunk_info->fspace, type_info->dst_type_size) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTINIT, FAIL, "unable to initialize file selection information") mem_iter_init = TRUE; - if ((iter_nelmts = H5S_GET_SELECT_NPOINTS(chunk_info->fspace)) < 0) + if((iter_nelmts = H5S_GET_SELECT_NPOINTS(chunk_info->fspace)) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTCOUNT, FAIL, "dataspace is invalid") /* Scatter the owner's modification data into the chunk data buffer according to @@ -3231,44 +3235,44 @@ H5D__filtered_collective_chunk_entry_io(H5D_filtered_collective_io_info_t *chunk if(H5D__scatter_mem(tmp_gath_buf, chunk_info->fspace, mem_iter, (size_t) iter_nelmts, chunk_entry->buf) < 0) HGOTO_ERROR(H5E_DATASET, H5E_READERROR, FAIL, "couldn't scatter to chunk data buffer") - if (H5S_SELECT_ITER_RELEASE(mem_iter) < 0) + if(H5S_SELECT_ITER_RELEASE(mem_iter) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTFREE, FAIL, "couldn't release selection iterator") mem_iter_init = FALSE; - if (MPI_SUCCESS != (mpi_code = MPI_Waitall(chunk_entry->async_info.num_receive_requests, + if(MPI_SUCCESS != (mpi_code = MPI_Waitall(chunk_entry->async_info.num_receive_requests, chunk_entry->async_info.receive_requests_array, MPI_STATUSES_IGNORE))) HMPI_GOTO_ERROR(FAIL, "MPI_Waitall failed", mpi_code) /* For each asynchronous receive call previously posted, receive the chunk modification * buffer from another rank and update the chunk data */ - for (i = 0; i < (size_t) chunk_entry->async_info.num_receive_requests; i++) { + for(i = 0; i < (size_t) chunk_entry->async_info.num_receive_requests; i++) { const unsigned char *mod_data_p; /* Decode the process' chunk file dataspace */ mod_data_p = chunk_entry->async_info.receive_buffer_array[i]; - if (NULL == (dataspace = H5S_decode(&mod_data_p))) + if(NULL == (dataspace = H5S_decode(&mod_data_p))) HGOTO_ERROR(H5E_DATASET, H5E_CANTDECODE, FAIL, "unable to decode dataspace") - if (H5S_select_iter_init(mem_iter, dataspace, type_info->dst_type_size) < 0) + if(H5S_select_iter_init(mem_iter, dataspace, type_info->dst_type_size) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTINIT, FAIL, "unable to initialize memory selection information") mem_iter_init = TRUE; - if ((iter_nelmts = H5S_GET_SELECT_NPOINTS(dataspace)) < 0) + if((iter_nelmts = H5S_GET_SELECT_NPOINTS(dataspace)) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTCOUNT, FAIL, "dataspace is invalid") /* Update the chunk data with the received modification data */ if(H5D__scatter_mem(mod_data_p, dataspace, mem_iter, (size_t) iter_nelmts, chunk_entry->buf) < 0) HGOTO_ERROR(H5E_DATASET, H5E_WRITEERROR, FAIL, "couldn't scatter to write buffer") - if (H5S_SELECT_ITER_RELEASE(mem_iter) < 0) + if(H5S_SELECT_ITER_RELEASE(mem_iter) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTFREE, FAIL, "couldn't release selection iterator") mem_iter_init = FALSE; - if (dataspace) { - if (H5S_close(dataspace) < 0) + if(dataspace) { + if(H5S_close(dataspace) < 0) HGOTO_ERROR(H5E_DATASPACE, H5E_CANTFREE, FAIL, "can't close dataspace") dataspace = NULL; - } + } /* end if */ H5MM_free(chunk_entry->async_info.receive_buffer_array[i]); } /* end for */ |