CGNS version 4.1.0
INTRODUCTION
This document describes the difference between CGNS 4.0.0 and
CGNS 4.1.0, and contains information on known problems in
CGNS 4.1.0.
Links to the CGNS current released source code can be found at:
User documentation for the current release can be found at:
http://cgns.org/CGNS_docs_current/midlevel/index.html
For more information, see the CGNS home page:
CONTENTS
- New Features
- Support for new platforms and languages
- Bug Fixes since CGNS 4.0.0
- Known Problems
New Features
Configuration:
-------------
Library:
--------
- Implemented CPEX42 (Storing bounding box of a grid, CGNS-149)
- Implemented CPEX43 (Family hierarchy as a tree, CGNS-180)
- Switch to using HDF5 Compact storage for smaller datasets
Added HDF5 compact storage to CGNS to improve parallel IO performance. The default storage
is compact storage unless the dataset does not meet the < 64KiB limit. Also, if the
dataset can have partial IO, then it is contiguous storage. Also, removed 'tab' spacing.
All changes should be transparent to the application code. (CGNS-160, PR-130)
Parallel Library:
-----------------
Fortran Library:
----------------
- Introduced Fortran wrappers for CPEX42 and CPEX43
Tools:
------
Bug Fixes since CGNS 4.0.0 release
Library:
-------
****************** REMOVED APIS **************************
Removed the APIs:
cgio_read_all_data, cgio_read_data, cgio_read_block_data
cgio_read_all_data_f, cgio_read_data_f, cgio_read_block_data_f
These APIs should be used instead:
cgio_read_all_data_type, cgio_read_data_type, cgio_read_block_data_type
cgio_read_all_data_type_f, cgio_read_data_type_f, cgio_read_block_data_type_f
(CGNS-192)
****************** REMOVED APIS **************************
* Fixed tests for scoped enum use. (PR-123)
* Unified similar defines. (PR-124)
* Fixed bad assert. (PR-125)
* Corrected print format. (CGNS-152, PR-126)
* (CGNS-158, PR #128)
Fixed heap buffer overflow in cgi_read_ptset
In cgi_read_ptset, there is made a difference between point set of list and range type.
The 'CellListDonor' type was missing in the list of list types, and this led to a heap buffer overflow.
Added 'CellListDonor' type to list of list types.
* (CGNS-159, PR #127)
Fixed global-buffer-overflow in ADFI_string_2_C_string.
Added a search loop for early NULL termination of the string passed to ADFI_string_2_C_string in
order to prevent global-buffer-overflow when string literals are given as a parameter.
* Fixed return stat after calling MPI_Finalize and switched to using MPI_AllReduce to get the same
exit code on all the processes.
* Fixed typo, zc should be z3
* Fixed Problems linking Fortran code with 3.3.1 shared library (w/ CMake) (CGNS-118, PR-131)
* Do not read elements when computing partial data size. (PR-137)
* Disabled the use of H5Pset_file_space_strategy as it is not compatible with HDF5 v1.8 (CGNS-166)
* Use correct MPI communicator (PR-140)
The `MPI_Allreduce` call in `cgp_parent_data_write()` is using `MPI_COMM_WORLD`, but it would be more correct
to use the communicator stored in `cgp_mpi_comm` which is set by the client. If the client is not using
`MPI_COMM_WORLD` and calls this function, it will hang since not all ranks will be participating in the call.
Expanded open-close test to include mixed serial and parallel open and close
* Fixed incorrect variable type (PR-143)
Minor fix; only applicable if more than 2.1 billion intervals in a range, but does eliminate a
compiler warning -- `npt` should be `cgsize_t`
* Fixed compiler warnings about the possible loss of data (PR-144)
A few variables defined as `int` should be `ssize_t` to avoid potential loss of data on large models.
* Corrupted memory when reading int32 connectivity into an int64 dataspace, (CGNS-157)
Fixed by changing the cgio_read* calls internal to the MLL to use the
cgio_read*_type APIs instead. Created a new cgio_read_block_data_type API with
an added data type parameter.
* Removed non-standard unlink call
* Make static mpi-related variables consistent (PR-139)
Currently, it is possible for the `pcg_mpi_comm` to be inconsistent with
`pcg_mpi_comm_rank` and `pcg_mpi_comm_size`. The latter two values are set during a
call to `cgp_open,` and the first is set during a call to `cgp_mpi_comm.` If the user
is using both parallel and serial file access during the same run, they may call
`cgp_mpi_comm` multiple times, and the values of `pcg_mpi_comm_rank` and `pcg_mpi_comm_size`
will only be consistent with the communicator in `pcg_mpi_comm` if the user calls `cgp_open.`
With this fix, all three values (and the `ParallelMPICommunicator`) will always be consistent.
Also, if the user does not call `cgp_mpi_comm()` before calling `cgp_open(),`
this will be detected, and everything will be consistent.
Initialize communicator to MPI_COMM_NULL
* miscellaneous code quality improvements and warning fixes.
Fortran:
--------
* Added missing BUILD_PARALLEL_F. (ref. PR-124)
* Fixed cgiof_f03.F90 test passing literal character instead of NULL character in CGIO_CREATE_LINK_F
Configuration:
-------------
* Fixed HDF5 CMake feature detection for parallel Collective_metadata and
H5Pset_file_space_strategy. (PR-138)
* Fixed HDF5 feature detection (PR-138)
Tools:
------
* Fix cgnstools install (PR-129)
Known Problems
* See https://cgnsorg.atlassian.net
Supported Platforms
The following platforms are supported and have been tested for this release.
They are built with the configure process unless specified otherwise.
Linux 2.6.32-696.16.1.el6.ppc64 gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-18)
#1 SMP ppc64 GNU/Linux g++ (GCC) 4.4.7 20120313 (Red Hat 4.4.7-18)
(ostrich) GNU Fortran (GCC) 4.4.7 20120313 (Red Hat 4.4.7-18)
IBM XL C/C++ V13.1
IBM XL Fortran V15.1
Linux 3.10.0-327.10.1.el7 GNU C (gcc), Fortran (gfortran), C++ (g++)
#1 SMP x86_64 GNU/Linux compilers:
(kituo/moohan) Version 4.8.5 20150623 (Red Hat 4.8.5-4)
Version 4.9.3, 5.2.0, 7.1.0
Intel(R) C (icc), C++ (icpc), Fortran (icc)
compilers:
Version 17.0.0.098 Build 20160721
MPICH 3.1.4
Linux-3.10.0- spectrum-mpi/rolling-release with cmake>3.10 and
862.14.4.1chaos.ch6.ppc64le clang/3.9,8.0
#1 SMP ppc64le GNU/Linux gcc/7.3
(ray) xl/2016,2019
Linux 3.10.0- openmpi/3.1,4.0 with cmake>3.10 and
957.12.2.1chaos.ch6.x86_64 clang 5.0
#1 SMP x86_64 GNU/Linux gcc/7.3,8.2
(serrano) intel/17.0,18.0/19.0
Linux 3.10.0- openmpi/3.1/4.0 with cmake>3.10 and
1062.1.1.1chaos.ch6.x86_64 clang/3.9,5.0,8.0
#1 SMP x86_64 GNU/Linux gcc/7.3,8.1,8.2
(chama,quartz) intel/16.0,18.0,19.0
Linux 4.4.180-94.100-default cray-mpich/7.7.6 with PrgEnv-*/6.0.5, cmake>3.10 and
#1 SMP x86_64 GNU/Linux gcc/7.2.0,8.2.0
(mutrino) intel/17.0,18.0
Linux 4.14.0- spectrum-mpi/rolling-release with cmake>3.10 and
49.18.1.bl6.ppc64le clang/6.0,8.0
#1 SMP ppc64le GNU/Linux gcc/7.3
(lassen) xl/2019
SunOS 5.11 32- and 64-bit Sun C 5.12 SunOS_sparc
(emu) Sun Fortran 95 8.6 SunOS_sparc
Sun C++ 5.12 SunOS_sparc
Windows 10 x64 Visual Studio 2015 w/ Intel Fortran 18 (cmake)
Visual Studio 2017 w/ Intel Fortran 19 (cmake)
Visual Studio 2019 w/ Intel Fortran 19 (cmake)
macOS 10.13.6 High Sierra Apple LLVM version 10.0.0 (clang/clang++-1000.10.44.4)
64-bit gfortran GNU Fortran (GCC) 6.3.0
(bear) Intel icc/icpc/ifort version 19.0.4
macOS 10.14.6 Mohave Apple LLVM version 10.0.1 (clang/clang++-1001.0.46.4)
64-bit gfortran GNU Fortran (GCC) 6.3.0
(bobcat) Intel icc/icpc/ifort version 19.0.4
Tested Configuration Features Summary
In the table below
y = tested
n = not tested in this release
x = not working in this release
| Platform | C | C [1] | Fortran | Fortran [1] |
|---|---|---|---|---|
| SunOS 5.11 32-bit | y | n | y | n |
| SunOS 5.11 64-bit | y | n | y | n |
| Windows 10 | y | n | n | n |
| Windows 10 x64 | y | n | n | n |
| Windows 10 Cygwin | n | n | x | n |
| Mac OS X 10.13.6 64-bit | y | n | y | n |
| Mac OS X 10.14.6 64-bit | y | n | y | n |
| CentOS 6.7 Linux 2.6.32 x86_64 GNU | y | y | y | y |
| CentOS 6.7 Linux 2.6.32 x86_64 Intel | y | y | y | y |
| CentOS 6.7 Linux 2.6.32 x86_64 PGI | y | n | y | n |
| CentOS 7.2 Linux 3.10.0 x86_64 GNU | y | y | y | y |
| CentOS 7.2 Linux 3.10.0 x86_64 Intel | y | y | y | y |
| Linux 2.6.32-431.11.2.el6.ppc64 | y | n | y | n |
[1] Parallel