diff options
author | Jordan Henderson <jhenderson@hdfgroup.org> | 2019-02-12 01:13:39 (GMT) |
---|---|---|
committer | Jordan Henderson <jhenderson@hdfgroup.org> | 2019-02-12 01:13:39 (GMT) |
commit | 34508f0620363d90ef3f76b314d52e4b01c20a81 (patch) | |
tree | af8054ab51fc5f4cc495882df9ed4343960586b9 /release_docs | |
parent | 65a820ae8981a84fe7fbac87c48482e9f82b35f4 (diff) | |
parent | fa83ab9f7c7dd7966f475932325b5e3740810cfd (diff) | |
download | hdf5-34508f0620363d90ef3f76b314d52e4b01c20a81.zip hdf5-34508f0620363d90ef3f76b314d52e4b01c20a81.tar.gz hdf5-34508f0620363d90ef3f76b314d52e4b01c20a81.tar.bz2 |
Merge in latest from develop
Diffstat (limited to 'release_docs')
-rw-r--r-- | release_docs/INSTALL_CMake.txt | 1 | ||||
-rw-r--r-- | release_docs/README_HPC | 79 | ||||
-rw-r--r-- | release_docs/RELEASE.txt | 34 |
3 files changed, 104 insertions, 10 deletions
diff --git a/release_docs/INSTALL_CMake.txt b/release_docs/INSTALL_CMake.txt index 26cf3ce..233c2cc 100644 --- a/release_docs/INSTALL_CMake.txt +++ b/release_docs/INSTALL_CMake.txt @@ -635,6 +635,7 @@ HDF5_ENABLE_DIRECT_VFD "Build the Direct I/O Virtual File Driver" HDF5_ENABLE_EMBEDDED_LIBINFO "embed library info into executables" ON HDF5_ENABLE_HSIZET "Enable datasets larger than memory" ON HDF5_ENABLE_PARALLEL "Enable parallel build (requires MPI)" OFF +HDF5_ENABLE_PREADWRITE "Use pread/pwrite in sec2/log/core VFDs in place of read/write (when available)" ON HDF5_ENABLE_TRACE "Enable API tracing capability" OFF HDF5_ENABLE_USING_MEMCHECKER "Indicate that a memory checker is used" OFF HDF5_GENERATE_HEADERS "Rebuild Generated Files" ON diff --git a/release_docs/README_HPC b/release_docs/README_HPC new file mode 100644 index 0000000..bdeab67 --- /dev/null +++ b/release_docs/README_HPC @@ -0,0 +1,79 @@ +HDF5 version 1.11.4 currently under development + +HDF5 source tar files with the HPC prefix are intended for use on clusters where +configuration and build steps will be done on a login node and executable and +lib files that are built will be run on compute nodes. + +Note these differences from the regular CMake tar and zip files: + - Test programs produced by this tar file will be run using batch scripts. + - Serial and parallel HDF5options.cmake files, using parallel options by default. + +Note also that options are now available in HDF5 source to facilitate use of +toolchain files for using cross compilers available on login nodes to compile +HDF5 for compute nodes. + +Instructions to configure build and test HDF5 using CMake: + +1. The cmake version must be 3.10 or later (cmake --version). +2. Load or switch modules and set CC, FC, CXX for compilers desired. +3. run build-unix.sh to configure, build, test and package HDF5 with CMake. + +Contents: + +build-unix.sh Simple script for running CMake to configure, build, + test, and package HDF5. +CTestScript.cmake CMake script to configure, build, test and package + HDF5. +hdf5-<version> HDF5 source for <version>. +HDF5config.cmake CMake script to configure, build, test and package + HDF5. +HDF5Examples Source for HDF5 Examples. +HDF5options.cmake symlink to parallel or serial HDF5options.cmake files. + Default is parallel file, which builds and tests both + serial and parallel C and Fortran wrappers. + To build serial only, C Fortran and C++ wrappers, delete + The HDF5options.cmake link and run + 'ln -s ser-HDF5options.cmake HDF5options.cmake' to switch. +par-HDF5options.cmake Options file for HDF5 serial and parallel build and test. +ser-HDF5options.cmake Options file for HDF5 serial only build and test. +SZip.tar.gz Source for building SZip. +ZLib.tar.gz Source for buildng Zlib. + + +To cross compile with this HPC-CMake tar.gz HDF5 source file: +On Cray XC40 haswell login node for knl compute nodes using CMake and Cray modules: + 1. Uncomment line in HDF5options.txt to use a toolchain file - line 106 for + config/toolchain/crayle.cmake. + 2. Uncomment lines 110, 111, and 115 - 122 of HDF5options.cmake. + Line 110 allows configuring to complete on the haswell node. + Line 111 switches the compiler to build files for knl nodes. + Lines 115 - 122 set up test files to use sbatch to run build tests + in batch jobs on a knl compute node with 6 processes. + 3. Compiler module may be the default PrgEnv-intel/6.0.4 to use + intel/18.0.2 or other intel, PrgEnv-cray/6.0.4 to use cce/8.7.4, + or PrgEnv-gnu/6.0.4 for GCC compilers. PrgEnv-pgi/6.0.4 is also + available but has not been tested with this tar file. + 4. These CMake options are set in config/toolchain/crayle.cmake: + set(CMAKE_SYSTEM_NAME Linux) + set(CMAKE_COMPILER_VENDOR "CrayLinuxEnvironment") + set(CMAKE_C_COMPILER cc) + set(CMAKE_CXX_COMPILER c++) + set(CMAKE_Fortran_COMPILER ftn) + set(CMAKE_CROSSCOMPILING_EMULATOR "") + + 5. Settings for two other cross-compiling options are also in the + config/toolchain files which do not seem to be necessary with the + Cray PrgEnv-* modules + a. HDF5_USE_PREGEN. This option, along with the HDF5_USE_PREGEN_DIR + CMake variable would allow the use of an appropriate H5Tinit.c + file with type information generated on a compute node to be used + when cross compiling for those compute nodes. The use of the + variables in lines 110 and 111 of HDF5options.cmake file seem to + preclude needing this option with the available Cray modules and + CMake options. + b. HDF5_BATCH_H5DETECT and associated CMake variables. This option + when properly configured will run H5detect in a batch job on a + compute node at the beginning of the CMake build process. It + was also found to be unnecessary with the available Cray modules + and CMake options. +- diff --git a/release_docs/RELEASE.txt b/release_docs/RELEASE.txt index 9e63045..ef3bda0 100644 --- a/release_docs/RELEASE.txt +++ b/release_docs/RELEASE.txt @@ -264,21 +264,21 @@ Bug Fixes since HDF5-1.10.3 release Library ------- - - Fix hangs with collective metadata reads during chunked dataset I/O + - Performance issue when closing an object - In the parallel library, it was discovered that when a particular - sequence of operations following a pattern of: + The slow down is due to the search of the "tag_list" to find + out the "corked" status of an object and "uncork" it if so. - "write to chunked dataset" -> "flush file" -> "read from dataset" + Improve porformance by skipping the search of the "tag_list" + if there are no "corked" objects when closing an object. - occurred with collective metadata reads enabled, hangs could be - observed due to certain MPI ranks not participating in the collective - metadata reads. + (VC - 2019/2/6) - To fix the issue, collective metadata reads are now disabled during - chunked dataset raw data I/O. + - Fixed a potential invalid memory access and failure that could occur when + decoding an unknown object header message (from a future version of the + library). - (JTH - 2019/02/11, HDFFV-10563, HDFFV-10688) + (NAF - 2019/01/07) - Deleting attributes in dense storage @@ -411,6 +411,11 @@ Bug Fixes since HDF5-1.10.3 release Testing ------- + - Fixed a test failure in testpar/t_dset.c caused by + the test trying to use the parallel filters feature + on MPI-2 implementations. + + (JTH, 2019/2/7) Bug Fixes since HDF5-1.10.2 release ================================== @@ -653,6 +658,15 @@ Bug Fixes since HDF5-1.10.2 release Testing ------- + - The dt_arith test failed on IBM Power8 and Power9 machines when testing + conversions from or to long double types, especially when special values + such as infinity or NAN were involved. In some cases the results differed + by extremely small amounts from those on other machines, while some other + tests resulted in segmentation faults. These conversion tests with long + double types have been disabled for ppc64 machines until the problems are + better understood and can be properly addressed. + + (SRL - 2019/01/07, TRILAB-98) Supported Platforms =================== |