From ed031ad89e257756630b0906912b084d189ea462 Mon Sep 17 00:00:00 2001 From: Daniel Garcia-Sanchez Date: Tue, 9 Oct 2018 16:37:55 +0200 Subject: [PATCH] Improve the documentation Thread safety documentation Modify the hyperslab documentation --- include/deal.II/base/hdf5.h | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/include/deal.II/base/hdf5.h b/include/deal.II/base/hdf5.h index 6b4c4bb7a5..8787327df1 100644 --- a/include/deal.II/base/hdf5.h +++ b/include/deal.II/base/hdf5.h @@ -127,7 +127,8 @@ DEAL_II_NAMESPACE_OPEN * ## Write a hyperslab in parallel * Hyperslabs are portions of datasets. A hyperslab can be a contiguous * collection of points in a dataset, or it can be a regular pattern of points - * or blocks in a datataset. + * or blocks in a datataset. Hyperslabs are equivalent to python numpy and h5py + * [slices](http://docs.h5py.org/en/latest/high/dataset.html#reading-writing-data). * * See the Dataspaces @@ -300,6 +301,14 @@ DEAL_II_NAMESPACE_OPEN * 4 5 * @endcode * + * # HDF5 and thread safety + * By default HDF5 is not thread-safe. The HDF5 library can be configured to be + * thread-safe, see [the HDF5 + * documentation](https://support.hdfgroup.org/HDF5/faq/threadsafe.html). The + * thread-safe HDF5 version serializes the API but does not provide any level of + * concurrency. To achieve high parallel performance with HDF5, we advice to use + * HDF5 with MPI. + * * @author Daniel Garcia-Sanchez, 2018 */ namespace HDF5 -- 2.39.5