diff options
| author | Mohamad Chaarawi <chaarawi@hdfgroup.org> | 2014-08-22 15:39:14 (GMT) |
|---|---|---|
| committer | Mohamad Chaarawi <chaarawi@hdfgroup.org> | 2014-08-22 15:39:14 (GMT) |
| commit | 91aab9734658ab0e7fc490d578d7945262281e48 (patch) | |
| tree | 076fc8380d43dc1e2e5e534c43910313859b4716 /release_docs | |
| parent | 2ef52022e0152586f004bbc2318f1e117ac8590e (diff) | |
| download | hdf5-91aab9734658ab0e7fc490d578d7945262281e48.zip hdf5-91aab9734658ab0e7fc490d578d7945262281e48.tar.gz hdf5-91aab9734658ab0e7fc490d578d7945262281e48.tar.bz2 | |
[svn-r25543] merge 25509 & 25524 from trunk.
Optimize chunked dataset fill operation by making it collective.
+ add a workaround for new algorithm to work with ompi.
Update RELEASE.txt with this new optimization.
Diffstat (limited to 'release_docs')
| -rw-r--r-- | release_docs/RELEASE.txt | 7 |
1 files changed, 6 insertions, 1 deletions
diff --git a/release_docs/RELEASE.txt b/release_docs/RELEASE.txt index 236fe0f..afa7a4f 100644 --- a/release_docs/RELEASE.txt +++ b/release_docs/RELEASE.txt @@ -67,7 +67,12 @@ New Features Parallel Library ---------------- - - + - Optimize the chunk fill algorithm of datasets by writing out all + the chunk fill values collectively in a single MPI-IO + call. Previously it was done independently by rank 0 + one block at a time. This should show a great performance + improvement when creating chunked datasets in parallel when the + chunk dimensions are fairly small. Tools ----- |