| Commit message (Collapse) | Author | Age |
| |
|
|
|
|
| |
compiler warnings
|
|
|
|
|
|
|
|
| |
Index HDF5 datafiles weree not handled correctly in the case of sliced
data. The files are created and initialized correctly at the first
iteration, but every subsequent access fails with an HDF5 error.
Patch by David Radice.
|
|
|
|
| |
also commented
|
| |
|
|
|
|
|
| |
Rewrite padding infrastructure.
Add padded array extents to transport operator APIs.
|
|
|
|
| |
only check datasets whose variables were actually requested to be read
|
|
|
|
| |
also garbage collect HDF5 at each H5close
|
| |
|
|
|
|
|
|
|
|
|
| |
rather than datasets of extent [1,1,1]. It so happens that this patch removes
the separate shape arrays and separate dataspaces for the datasets in index
files and creates the index file datasets with the same dataspace as the
"heavy" file datasets. It retains the h5shape attribute that was originally
introduced for index files even though it is now redundant (since one could
call H5Sget_simple_extent_dims on the datasets in the index files).
|
| |
|
| |
|
| |
|
|
|
|
| |
garbage collect HDF5 objects when closing files
|
|
|
|
|
|
| |
we try for index files when opening the first file and if this does not
succeed do not try opening index files again. This reduces the number of
file system accesses when no index files are present.
|
| |
|
|
|
|
| |
dataset
|
|
|
|
| |
which happens to be the output method used for single process runs.
|
|
|
|
| |
open_one_input_file_at_a_time
|
|
|
|
| |
with open_one_input_file_at_a_time
|
|
|
|
|
|
| |
to checkpoint when cctk_iteration % checkpoint_every_divisor == 0 rather
than at whenever checkpoint_every iterations have passed since the last
checkpoint
|
|
|
|
|
| |
once per variable, refinement level and time level that is. So still
about ~3000 warnings in a typical simulation.
|
|
|
|
|
|
|
| |
Move MPI support from flesh to thorn ExternalLibraries/MPI. This also
requires thorns that call MPI directly to declare this in their
configuration.ccl. Existing configurations using MPI need to include
ExternalLibraries/MPI into their thorn list.
|
|
|
|
| |
Right now there is no facility to actually use this test unfortunately.
|
|
|
|
|
|
|
| |
NOTE: this assumes (like other parts of CarpetIOHDF5) that the number of
symmetry points is the number of ghost points.
NOTE: it likely outputs too many points when RotatingSymmetry is used and only
buffer points touch the symmetry boundary.
|
|
|
|
| |
to disk
|
| |
|
|
|
|
|
|
|
| |
There are still systems with only version 1.6 installed, or broken installs
of version 1.8 (Debian system packages at least up to squeeze). These
systems benefit from a compiling hdf5_recombiner, while version 1.8 is
not really required to be used here.
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
Change the API to obtain a pointer to grid function data:
- Use a function "typed_data_pointer" instead of overloading the ()
operator (because this looks nicer)
- Don't use a virtual function (because this isn't needed)
- Update all uses
|
|
|
|
|
|
|
|
|
|
|
| |
computing coordinates of points
this is the same issue (just seems from the other side of the output) as in
"CarpetIOHDF5: Correct iorigin attribute for 2D output", namely that iorigin
is stored in multiples of the stride for the given refinement level
---
Carpet/CarpetIOHDF5/src/util/hdf5toascii_slicer.cc | 6 +++---
1 files changed, 3 insertions(+), 3 deletions(-)
|
|
|
|
|
|
|
|
| |
file descriptors
---
Carpet/CarpetIOHDF5/src/util/hdf5_slicer.cc | 13 +++++--------
1 files changed, 5 insertions(+), 8 deletions(-)
|
|
|
|
| |
is used by VisIt
|
|\ |
|
| | |
|
| |
| |
| |
| |
| |
| |
| | |
Introduce a new API to checkpoint only a subset of group, via an
aliased function IO_SetCheckpointGroups. This can be used for
simulation spawning, i.e. off-loading certain calculations (e.g.
analysis) outside of the main simulation.
|
| |
| |
| |
| |
| | |
IOUtil_DefaultIORequest returns a fresh copy of the default IO request,
rather than just a pointer to it. It has to be freeed afterwards.
|
| | |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
current request does not exist for one_file_per_group output
this is re #410:
--8<-- by Erik Schnetter --8<--
The corresponding code in CarpetIOHDF5.cc, which outputs data that are not
slices, uses the same algorithm. However, it contains an additional check
ensuring that a default request is used if the corresponding request does not
exist. Look for calls to IOUtil_DefaultIORequest to find this code. I believe
that an equivalent logic would correct this problem in OutputSlice.cc.
--8<-- by Erik Schnetter --8<--
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
---
Carpet/CarpetIOASCII/doc/documentation.tex | 2 +-
Carpet/CarpetIOHDF5/doc/documentation.tex | 2 +-
Carpet/CarpetInterp/doc/documentation.tex | 2 +-
Carpet/CarpetReduce/doc/documentation.tex | 2 +-
4 files changed, 4 insertions(+), 4 deletions(-)
|
| | |
|
|/
|
|
|
|
|
|
| |
Currently CarpetIOHDF5 uses the c compiler options with the c++ compiler
to build the utilities. This can lead to warnings about flags not
supported by c++ (only by c).
This patch let's it use the c++ compiler options instead.
|
|
|
|
|
|
| |
CarpetIOHDF5 already prints the iteration and time for periodic
checkpoints. This patch adds this to both of initial data checkpoints
(also after restart) and termination checkpoints.
|
|
|
|
|
|
| |
map number
this hopefully fixes ticket #446
|
|
|
|
| |
* added new cp/recovery par files for cell-centered case
|
| |
|
|
|
|
|
| |
Introduce a new parameter skip_recover_variables that skips recovery
on a set of variables.
|
|
|
|
|
|
|
|
| |
Specifically, remove any hierarchy information that has been added to
the name of timers, as well as any code for creating timers
dynamically, as these are now unnecessary. Additionally, time some
previously-untimed parts of the code and make timer names in some
places more consistent.
|
| |
|