diff options
author | Quincey Koziol <koziol@hdfgroup.org> | 2009-07-02 22:34:20 (GMT) |
---|---|---|
committer | Quincey Koziol <koziol@hdfgroup.org> | 2009-07-02 22:34:20 (GMT) |
commit | c243e1e13456365b0b555a9c82d5cc17f1436727 (patch) | |
tree | acf5130fd8bede482315c050b1ec0e3faf096ee4 /src/H5Dmpio.c | |
parent | 47e4e67e3cd11f99ac417613eed9f47e65f55468 (diff) | |
download | hdf5-c243e1e13456365b0b555a9c82d5cc17f1436727.zip hdf5-c243e1e13456365b0b555a9c82d5cc17f1436727.tar.gz hdf5-c243e1e13456365b0b555a9c82d5cc17f1436727.tar.bz2 |
[svn-r17148] Description:
Refactor how chunked dataset information is computed, moving it earlier and
avoiding more recomputation.
Tested on:
FreeBSD/32 6.3 (duty) in debug mode
FreeBSD/64 6.3 (liberty) w/C++ & FORTRAN, in debug mode
Linux/32 2.6 (jam) w/PGI compilers, w/C++ & FORTRAN, w/threadsafe,
in debug mode
Linux/64-amd64 2.6 (smirom) w/Intel compilers w/default API=1.6.x,
w/C++ & FORTRAN, in production mode
Solaris/32 2.10 (linew) w/deprecated symbols disabled, w/C++ & FORTRAN,
w/szip filter, in production mode
Linux/64-ia64 2.6 (cobalt) w/Intel compilers, w/C++ & FORTRAN,
in production mode
Linux/64-ia64 2.4 (tg-login3) w/parallel, w/FORTRAN, in debug mode
Linux/64-amd64 2.6 (abe) w/parallel, w/FORTRAN, in production mode
Mac OS X/32 10.5.7 (amazon) in debug mode
Mac OS X/32 10.5.7 (amazon) w/C++ & FORTRAN, w/threadsafe,
in production mode
Diffstat (limited to 'src/H5Dmpio.c')
-rw-r--r-- | src/H5Dmpio.c | 20 |
1 files changed, 10 insertions, 10 deletions
diff --git a/src/H5Dmpio.c b/src/H5Dmpio.c index afa9201..e111404 100644 --- a/src/H5Dmpio.c +++ b/src/H5Dmpio.c @@ -843,7 +843,7 @@ H5D_link_chunk_collective_io(H5D_io_info_t *io_info, const H5D_type_info_t *type } /* end if */ /* Retrieve total # of chunks in dataset */ - H5_ASSIGN_OVERFLOW(total_chunks, fm->total_chunks, hsize_t, size_t); + H5_ASSIGN_OVERFLOW(total_chunks, fm->layout->u.chunk.nchunks, hsize_t, size_t); /* Handle special case when dataspace dimensions only allow one chunk in * the dataset. [This sometimes is used by developers who want the @@ -997,7 +997,7 @@ if(H5DEBUG(D)) total_chunk_addr_array = H5MM_malloc(sizeof(haddr_t) * total_chunks); /* Retrieve chunk address map */ - if(H5D_chunk_addrmap(io_info, total_chunk_addr_array, fm->down_chunks) < 0) + if(H5D_chunk_addrmap(io_info, total_chunk_addr_array) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTGET, FAIL, "can't get chunk address") /* Get chunk with lowest address */ @@ -1103,7 +1103,7 @@ H5D_multi_chunk_collective_io(H5D_io_info_t *io_info, const H5D_type_info_t *typ #endif /* Retrieve total # of chunks in dataset */ - H5_ASSIGN_OVERFLOW(total_chunk, fm->total_chunks, hsize_t, size_t); + H5_ASSIGN_OVERFLOW(total_chunk, fm->layout->u.chunk.nchunks, hsize_t, size_t); HDassert(total_chunk != 0); /* Allocate memories */ @@ -1706,7 +1706,7 @@ H5D_sort_chunk(H5D_io_info_t *io_info, const H5D_chunk_map_t *fm, * 0, we would always want to obtain the chunk addresses individually * for each process. */ - bsearch_coll_chunk_threshold = (sum_chunk * 100) / ((int)fm->total_chunks * mpi_size); + bsearch_coll_chunk_threshold = (sum_chunk * 100) / ((int)fm->layout->u.chunk.nchunks * mpi_size); if((bsearch_coll_chunk_threshold > H5D_ALL_CHUNK_ADDR_THRES_COL) && ((sum_chunk / mpi_size) >= H5D_ALL_CHUNK_ADDR_THRES_COL_NUM)) many_chunk_opt = H5D_OBTAIN_ALL_CHUNK_ADDR_COL; @@ -1725,19 +1725,19 @@ if(H5DEBUG(D)) HDfprintf(H5DEBUG(D), "Coming inside H5D_OBTAIN_ALL_CHUNK_ADDR_COL\n"); #endif /* Allocate array for chunk addresses */ - if(NULL == (total_chunk_addr_array = H5MM_malloc(sizeof(haddr_t) * (size_t)fm->total_chunks))) + if(NULL == (total_chunk_addr_array = H5MM_malloc(sizeof(haddr_t) * (size_t)fm->layout->u.chunk.nchunks))) HGOTO_ERROR(H5E_RESOURCE, H5E_NOSPACE, FAIL, "unable to allocate memory chunk address array") /* Retrieve all the chunk addresses with process 0 */ if((mpi_rank = H5F_mpi_get_rank(io_info->dset->oloc.file)) < 0) HGOTO_ERROR(H5E_IO, H5E_MPI, FAIL, "unable to obtain mpi rank") if(mpi_rank == 0) { - if(H5D_chunk_addrmap(io_info, total_chunk_addr_array, fm->down_chunks) < 0) + if(H5D_chunk_addrmap(io_info, total_chunk_addr_array) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTGET, FAIL, "can't get chunk address") } /* end if */ /* 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->total_chunks), MPI_BYTE, (int)0, io_info->comm))) + 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) } /* end if */ @@ -1869,12 +1869,12 @@ H5D_obtain_mpio_mode(H5D_io_info_t* io_info, H5D_chunk_map_t *fm, HGOTO_ERROR(H5E_IO, H5E_MPI, FAIL, "unable to obtain mpi size") /* Setup parameters */ - H5_ASSIGN_OVERFLOW(total_chunks, fm->total_chunks, hsize_t, int); + H5_ASSIGN_OVERFLOW(total_chunks, fm->layout->u.chunk.nchunks, hsize_t, int); percent_nproc_per_chunk = H5P_peek_unsigned(dx_plist, H5D_XFER_MPIO_CHUNK_OPT_RATIO_NAME); #if defined(H5_MPI_COMPLEX_DERIVED_DATATYPE_WORKS) && defined(H5_MPI_SPECIAL_COLLECTIVE_IO_WORKS) chunk_opt_mode = (H5FD_mpio_chunk_opt_t)H5P_peek_unsigned(dx_plist, H5D_XFER_MPIO_CHUNK_OPT_HARD_NAME); if((chunk_opt_mode == H5FD_MPIO_CHUNK_MULTI_IO) || (percent_nproc_per_chunk == 0)) { - if(H5D_chunk_addrmap(io_info, chunk_addr, fm->down_chunks) < 0) + if(H5D_chunk_addrmap(io_info, chunk_addr) < 0) HGOTO_ERROR(H5E_DATASET, H5E_CANTGET, FAIL, "can't get chunk address"); for(ic = 0; ic < total_chunks; ic++) assign_io_mode[ic] = H5D_CHUNK_IO_MODE_COL; @@ -1931,7 +1931,7 @@ H5D_obtain_mpio_mode(H5D_io_info_t* io_info, H5D_chunk_map_t *fm, #endif /* calculating the chunk address */ - if(H5D_chunk_addrmap(io_info, chunk_addr, fm->down_chunks) < 0) { + if(H5D_chunk_addrmap(io_info, chunk_addr) < 0) { HDfree(nproc_per_chunk); #if !defined(H5_MPI_COMPLEX_DERIVED_DATATYPE_WORKS) || !defined(H5_MPI_SPECIAL_COLLECTIVE_IO_WORKS) HDfree(ind_this_chunk); |