eb2f4c6a01
Added NEWS stub for new version.
2019-05-16 21:36:26 +01:00
c5c43c4f3c
Changed the version name to alpha to differentiate. Now mdtest has the version number from IOR.
2019-05-16 21:25:34 +01:00
3a873ea2b0
Fix #150 .
2019-05-16 19:35:00 +01:00
6abdcc363a
Trivial fix for #149
2019-05-16 12:56:56 +01:00
9ead56c24e
Merge pull request #148 from hpc/fix-146
...
Moving options to submodules + fixes #146 + fixes #144
2019-05-14 14:32:37 -07:00
30a1ca187b
Bugfix by rmn1 for #147 and #112 .
2019-05-14 14:55:11 +01:00
54522315ec
Fix #146 ; handling of hints. Why are GPFS hints special, remains unclear.
2019-05-08 14:02:40 +01:00
a9b0bf35bc
Support MDTest verbosity >=5 prints behavior for all ranks instead of just rank 0 (which is enabled with verbosity level 3 as before)
2019-05-02 09:58:35 +01:00
6732fd2370
Merge pull request #145 from ThinkParQ/master
...
beegfs: don't fail if output exists
2019-04-30 09:57:48 -07:00
e3afdb5ed5
beegfs: don't fail if output exists
...
Use existing output file if it has the desired stripe settings
(number of targets, chunk size)
2019-04-29 15:52:26 +02:00
8fd755147a
Fix for #144
2019-04-19 10:20:04 +01:00
27ac1b88af
Removed verbose output.
2019-04-02 12:34:35 +01:00
07217c93aa
Bugfix for the parsing of flags. Backwards compatibility for direct IO.
2019-04-02 12:08:31 +01:00
21405ed924
Updated MMap module to option parser.
2019-03-27 22:32:59 +00:00
b4fb80a69e
Allowed option parser to reset flags (needed in scriptfiles).
2019-03-27 22:32:38 +00:00
07ca5247d7
Checked proper IOR behavior for -f option.
2019-03-27 20:37:46 +00:00
20e960d020
Support the setting of module-specific options per test.
2019-03-27 20:04:48 +00:00
d49d345007
Merge pull request #137 from DDNStorage/memoryPerNode
...
Fix segfault with option memoryPerNode
2019-03-14 18:33:24 +00:00
fabb1f4d3b
Fix segfault with option memoryPerNode
...
optarg contains a NULL address, which causes a segfault in
NodeMemoryStringToBytes().
2019-03-14 19:20:18 +01:00
b22a39887a
Merge pull request #133 from DDNStorage/master
...
Enable IME native backend with mdtest. As Glenn also found it OK, I do too, there is a minor issue with the aiori backend, but well. For later.
2019-03-11 14:30:24 +00:00
110153cf3a
mdtest: Enable IME native interface
...
This patch makes mdtest allow IME native interface.
It also defines a generic way to enable mdtest with each
backend.
2019-02-12 15:50:51 +01:00
c6abf85378
aiori-IME: Add support of statfs in IME backend
...
This patch adds the support of statfs in the IME backend.
It uses ime_native_statvfs call to fill the ior_aiori_statfs_t
structure.
2019-02-11 15:00:18 +01:00
4c1c73ecae
add malloc check to fix
2019-01-23 16:02:58 -08:00
20bb99e1b7
Potential fix for #127 => Segmentation fault due to memory protection.
2019-01-19 17:00:49 +00:00
e71cded6c3
Merge pull request #124 from hpc/fix-hdf5
...
Bugfix for HDF5 plugin; readded collective_md.
2019-01-02 10:01:56 -08:00
a9385f4e11
Merge pull request #123 from hpc/fix-112
...
Fix 112 + new feature to compare patterns.
2019-01-02 10:01:26 -08:00
1dbca5c293
Bugfix for HDF5 plugin; readded collective_md.
2018-12-23 13:07:23 +00:00
4d3d06da7a
New testing framework for mdtest. The framework compares the observed with the expected pattern for the runs. For a new test, the pattern is recorded. Note that the tests are enumerated.
2018-12-20 21:09:34 +00:00
ed421bcc7b
Fixes that issue. Hopefully works for all combinations.
2018-12-20 21:08:55 +00:00
dced60b060
Added test for bug #112 .
2018-12-20 20:40:53 +00:00
1b28a64234
Merge pull request #120 from hpc/fix-119
...
Reset stonewall timer to make it work again with running all phases
2018-12-19 15:09:05 -08:00
7b603bc5a9
Reset stonewall timer to make it work again with running all phases in one execution; i.e. not using multiple runs specifying: -C, -r
2018-12-19 21:37:37 +00:00
4297ccd7f2
Bugfix for stonewall computation.
2018-12-19 20:29:47 +00:00
741031312d
Merge pull request #116 from adilger/b_lustre_user
...
lustre: fix checks for lustre_user.h header
2018-12-14 16:17:24 -08:00
1f217ae48c
lustre: fix checks for lustre_user.h header
...
Fix the configure check if --with-lustre is specified, but
the linux/lustre/lustre_user.h header is not present. Only
one of the headers needs to be included if both are found.
In some cases, FASYNC is not defined, but forms part of the
O_LOV_DELAY_CREATE value, add a #define in that case.
Fixes #115 .
Fixes: cb88c4c19a831d94b864c49a162e2635730540e5
Signed-off-by: Andreas Dilger <adilger@dilger.ca>
2018-12-14 16:45:43 -07:00
c9151d7495
merge RC for 3.2 back into master
2018-12-04 13:33:11 -08:00
11469ac5a9
updated NEWS for final contributions to 3.2 release
2018-12-04 13:08:07 -08:00
e1158daaf2
Merge pull request #105 from hpc/feature-test-scripts-85
...
Dry-run option to test the configuration / scripts.
2018-12-04 15:06:47 -06:00
e045c6024e
Reverted the change to HDF5 module back.
2018-12-04 12:59:17 -08:00
4aada0a32f
Updated docu for the changed behavior.
2018-12-04 12:59:12 -08:00
fec0645c04
Changed the parser.
...
The parser now supports concurrent parsing of all plugin options.
Moved HDF5 collective_md option into the backend as an example.
Example: ./src/ior -a dummy --dummy.delay-xfer=50000
2018-12-04 12:58:47 -08:00
efaa0c411a
HDF5: option to enable collective metadata
...
In HDF5-1.10 the library has an option to read and write the metadata of a file collectively. One has to opt-in to this optimization with a property list but particularly at large scale it can have tremendous performance implications.
Further information:
- https://support.hdfgroup.org/HDF5/docNewFeatures/CollMDataIO/RFC-CollectiveMetadataReads.pdf
- https://support.hdfgroup.org/HDF5/doc/RM/RM_H5P.html#Property-SetAllCollMetadataOps
2018-12-04 12:54:05 -08:00
5e2e1cc96f
Merge pull request #104 from hpc/feature-hdf5-backend
...
Changed the parser to fix #98 . Does _not_ yet have the HDF5 backend args updated. They were reverted to allow this to make the 3.2 release and should be re-applied as a separate commit.
2018-12-04 14:41:31 -06:00
5c847f552a
Merge pull request #110 from hpc/fix-gnu99-101
...
Moved _XOPEN_SOURCE to config.h; resolves #93
2018-12-04 14:13:16 -06:00
7db8ffd8ab
Reverted the change to HDF5 module back.
2018-12-04 19:16:41 +00:00
71fb1ceae8
Merge pull request #113 from wangshilong/master
...
mdtest: fix calculation for barriers
2018-11-24 11:53:21 +00:00
8902bc8722
mdtest: fix calculation for barriers
...
Currently, calculation for barriers case looks a bit
suprising, for example, Min stats is from min of
Max value of number of mpi proc, this makes Min
and Max very similar in our testing thus we get
a small Std value too.
I am not a MPI expert, but question is why we
don't use more nature calcuation which calculate
all results from different iterations and MPI proc?
Signed-off-by: Wang Shilong <wshilong@ddn.com>
2018-11-22 19:57:32 +08:00
5a8071fbd8
GNU99 extracted _XOPEN_SOURCE to config.h.
2018-10-28 18:34:22 +00:00
3a4f2ea33e
Merge pull request #101 from glennklockwood/issue93-squashed
...
compile correctly with -std=c99
2018-10-28 18:16:52 +00:00
f49a7e32fc
Bugfix #106 mostly.
2018-10-19 11:07:26 -07:00
Julian M. Kunkel