It shares the create/open/delete/set_version/get_file_size
functions with POSIX backend.
The mmap backend also supports fsync and fsyncPerWrite options,
and it will use msync() instead and fsync().
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Context: Some file systems require a prefix in the path.
The POSIX 'access' call fails and consequently files are
never deleted.
This patch implements an access function in the MPIIO
backend using MPI_File_open. Prefixes can now be parsed
by ROMIO.
This adds a .travis.yaml file to use travis-ci for continuous
integration.
It currently compiles IOR on an ubuntu 14.04 machine with
only the POSIX, MPIIO and HDF5 backends enabled. For Lustre, GPFS,
NCMPI, HDFS and S3 libraries needs to be compiled as they are not
present in the ubuntu repositories. Lustre and NCMPI should be easy when
ubuntu 16.04 is available at travis-ci. S3 could be compiled by source
which shouldn't be hard but needs to be done if one is interested.
Hadoop needs a more complex installation process. GPFS seams not
possible as it is proprietary and needs a license as far as i know.
Repo Owner needs to enable CI on travis-ci.org. It's free for open
source git hub repos.
The O_DIRECT option was not working as set_o_direct_flag() were moved to
utilities.c but there the #define _GNU_SOURCE where missing. This lead
to not the Waring "cannot use O_DIRECT".
Previously, the configure script could fail to find an MPI
implementation but continue regardless. This alters the behaviour so
that the configure script fails if the check for MPI does.
From Pnetcdf 1.7, the MPI datatype corresponding to NC_BYTE is change from MPI_BYTE to MPI_SIGNED_CHAR. If not change, running IOR with NCMPI will cause a fatal error as below:
ERROR in aiori-NCMPI.c (line 287): cannot write to data set.
ERROR: NetCDF: Not a valid data type or _FillValue type mismatch.
The file "src/aiori-NCMPI.c" uses numTasksWorld as the process which is declared in "src/ior.h". In "src/mdtest.c", NCMPI backend will be called if IOR configured with ncmpi enabled but numTasksWorld was not defined in "src/mdtest.c". So it will cause a compiler error like below:
mdtest-aiori-NCMPI.o: In function `NCMPI_Xfer':
/home/parallels/Documents/ior/src/aiori-NCMPI.c:272: undefined reference to `numTasksWorld'
Once a process hits the stonewall (timelimit), they all figure out the maximum pair read/written.
Each proces continues to read/write until the maximum number of pairs is reached, this simulates the wear out.
Mohamad Chaarawi