| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
| |
which can be set by passing to H5Fcreate and H5Fopen
a string of characters (one character for each flag to turn on)
as the value of MPI_Info with key H5F_MPIO_DEBUG_KEY.
|
| | |
|
| |
|
|
| |
for distribution.
|
| |
|
|
| |
to purge the file.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
./src/H5F.c
./src/HFcore.c
./src/H5Ffamily.c
./src/H5Flow.c
./src/H5Fmpio.c
./src/H5Fprivate.h
./src/H5Fsec2.c
./src/H5Fsplit.c
./src/H5Fstdio.c
./src/H5MF.c
./src/H5P.c
./src/H5Ppublic.h
Added H5Pset_alignment() so that it is now possible to align
file allocation requests on application-specified
boundaries. Any request >= the specified threshold will begin
on an address which is a multiple of the specified
alignment. Defaults are one for threshold and alignment. The
alignment is done on relative addresses, so the size of the
user block can affect the location of the data in the file.
./src/H5D.c
./src/dsets.c
Added a test for, and fixed the data space caching bug in
datasets. Extending a dataset through one handle will cause
all other handles to the same dataset to get the new dataset
size.
./src/H5S.c
./src/H5Sprivate.h
Removed an unused argument from H5S_read() which duplicated
information from the other argument.
./config/linux
Made `--enable-parallel' the default on my system. It used to
be that way before but then I accidently turned it off and
forgot about it.
./src/H5Fmpio.c
Qualified some function arguments with __unused__. Changed a
couple places where NULL was returned on error for herr_t
functions.
./src/H5P.c
Removed unused autos from H5Pset_mpi().
|
| |
|
|
|
|
| |
Made `--enable-parallel' the default on my system. It used to
be that way before but then I accidently turned it off and
forgot about it.
|
| |
|
|
|
| |
or parallel configurations.
Platform tested: O2K for both serial and parallel versions.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
H5Fmpio.c was not part of the serial distribution. The auto-
dependence configuration runs for serail distribution only, thus
H5Fmpio.c has no dependence. It ended up not recompiled even
when header files it depends on are changed.
Solution:
Changed configuration to always include this file.
Enclosed the whole code in a big #ifdef HAVE_PARALLEL macro
so that they would be compiled for parallel distribution only.
Platform tested:
O2K
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Parallel File open has a racing condition if file already exists.
Some processes see file exists, then truncate it (some system
by remove and create). Some processes does not see the file
exist, then attempt to just create it. Inconsistency.
Solution:
Instead of opening the file with COMM_SELF (which results in
racing condition in routine that calls it), open it with the
communicator in access_parms. (This assumes this access call
must be called collectively.)
Platform tested:
O2K.
|
| |
|
|
| |
to MPI functions. (Added by Kim Yates).
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
mode to use for data transfer. Currently, INDEPENDENT transfer
is the default mode. Collective is supported for files opened
via the MPIO driver. Specific changes to individual files follow.
H5D.c:
Replaced previous kludge of passing xfer_mode via the file access_mode
with xfer_mode function parameter all the way to the low level I/O
drivers.
H5Dprivate.h:
The H5D_xfer_t.xfer_mode is now used in serial and parallel modes.
Removed the "HAVE_PARALLEL" macro condition to support it.
H5Dpublic.h:
Added H5D_XFER_DFLT as a new value for H5D_transfer_t and removed
the HAVE_PARALLEL condition macro to make the H5D_transfer_t
definition available to both serial and parallel version.
H5B.c:
H5F.c:
H5Farray.c:
H5Fcore.c:
H5Ffamily.c:
H5Fistore.c:
H5Flow.c:
H5Fmpio.c:
H5Fsec2.c:
H5Fsplit.c:
H5Fstdio.c:
H5Gnode.c:
H5HG.c:
H5HL.c:
H5O.c:
H5Sprivate.h:
H5Ssimp.c:
debug.c:
Added xfer_mode to the argument lists of the following routines to
specify data transfer mode. Then updated calls to these routines
with appropriate transfer mode. Whenever cases that collective
mode is inappropriate (e.g. object header read-write, flush),
H5D_XFER_DFLT is used. The H5D_XFER_DFLT mode is currently
acting the same as the independent mode.
H5F_block_read
H5F_block_write
H5F_arr_read
H5F_arr_write
H5F_core_read
H5F_core_write
H5F_fam_read
H5F_fam_write
H5F_mpio_read
H5F_mpio_write
H5F_low_write
H5F_low_read
H5F_sec2_read
H5F_sec2_write
H5F_split_read
H5F_split_write
H5F_stdio_read
H5F_stdio_write
H5S_simp_fgath
H5S_simp_fscat
H5S_simp_read
H5S_simp_write
H5Fprivate.h:
H5P.c:
Removed access_mode field from struct H5F_access_t type.
The access-mode was the initial design to control independent
or collective I/O mode. It is now supported by the data
transfer mode.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
----------------------
./src/H5A.c
Named data types can have attributes.
Fixed bugs where the API functions didn't check the return
values of their internal counterparts and thus the automatic
error reporting didn't work.
Fixed some places where the error stack wasn't cleared after a
function returned failure.
Data types returned by H5Aget_type() are always read-only.
If the `attr_num' argument of H5Aiterate() is null then it
acts like H5Giterate() instead of failing -- it begins
processing attributes with the first one.
./src/H5D.c
We check for allocation overruns when scalar datasets are
stored in external files.
./src/H5O.c
H5O_modify() will fail if the message is >=16kB.
./src/H5Oattr.c
Split some long lines
./src/H5T.c
./src/H5Tprivate.h
Added H5T_entof() to support attributes on named types.
./src/h5ls.c
Prints the names of attributes and their sizes.
./test/cmpd_dset.c
./test/dsets.c
./test/dtypes.c
./test/extend.c
./test/external.c
./test/gheap.c
./test/istore.c
./test/links.c
./test/shtype.c
If the environment variable HDF5_NOCLEANUP is defined then the
temporary files are not removed. The testhdf5 program still
has the bug that it removes *.h5, clobbering test files from
other programs... oh well.
./test/dtypes.c
Added attribute tests.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
----------------------
./html/Datatypes.html
Updated documentation for named data types and type sharing.
./src/H5Farray.c
Split a couple long lines.
./src/H5T.c
The H5T_copy() demotes immutable types to read-only types so
they're cleaned up properly and memory is not leaked.
./test/dtypes.c
Modified to be more consistent with other tests. Removed all
internal header files, constants, types, and functions calls
since the API is now complete enough to test
everything. Temporary files are now removed.
|
| |
|
|
|
|
|
|
|
|
|
| |
Added tests for transient and named types and their use in
datasets.
./test/shtype.c
Commented out all the tests since they no longer apply.
./test/chunk.c
Removed a couple int->double coercion warnings.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
./src/H5D.c
./src/H5Tconv.c
./src/H5detect.c
Updated to work with new internal H5T functions. Fixed some
data type memory leaks during error recovery.
./src/H5Dprivate.h
Added H5D_typeof() similar to H5D_entof() that returns a
pointer directly to the dataset's type. This is used by
H5Tcopy() when invoked on a dataset (see below).
./src/H5Epublic.h
Fixed typos in H5E_BEGIN_TRY and H5E_END_TRY macros.
./src/H5F.c
Closing a file with objects still open reports the file name
in the warning message. Removed unnecessary invalidation of
shared data types.
./src/H5Gent.c
./src/H5Gpkg.h
./src/H5Gprivate.h
Added `const' to some arguments.
./src/H5O.c
./src/H5Oprivate.h
./src/H5Oshared.c
An object header message can now be a pointer to a message in
some other object header. The pointer is to the first message
of the appropriate type in the other object header and hard
link counts are maintained automatically to prevent dangling
pointers. The old global heap method of message sharing is
also supported although nothing actually uses it.
./src/H5Oattr.c
./src/H5Ocomp.c
./src/H5Ocont.c
./src/H5Odtype.c
./src/H5Oefl.c
./src/H5Olayout.c
./src/H5Oname.c
./src/H5Osdspace.c
./src/H5Oshare.c
./src/H5Ostab.c
Changed the data type for the shared message info struct to
H5O_shared_t and added an extra initializer to the class
methods struct for the set_share method.
./src/H5Odtype.c
Added the ability to share data type messages by pointing to
other object headers.
./src/H5T.c
./src/H5Tpkg.h
./src/H5Tprivate.h
./src/H5Tpublic.h
Added named data types and changed the functionality of some
API functions as follows:
* The term `read-only' means that a type handle cannot be
modified with functions like H5Tset_*() or H5Tinsert().
* The term `immutable' means the same as `read-only' with the
added restriction that H5Tclose() cannot be called for the
type. A transient type is made immutable by calling
H5Tlock().
* Handles to named types are always read-only.
* Handles to predefined types are immutable.
* A transient type can be converted to a named type by calling
H5Tcommit(). This function will fail if the type is already
named or is immutable.
* The API function H5Tcommitted() returns an indication of
whether a data type has been commited (or is named). If
H5Tcommitted() returns TRUE for a data type obtained by
calling H5Dget_type() on a dataset, then the dataset is
using a shared data type.
* H5Topen() returns a handle to a named type.
* H5Tcopy() always returns a handle to a modifiable transient
type even when invoked on a named type. Also, when invoked
on a dataset it returns a modifiable transient type which is
a copy of the dataset's data type.
* Using a named data type in the call to H5Dcreate() causes
the dataset object header to point to the named data type,
but using a transient type causes the type to be copied into
the dataset's object header.
* The data type returned from H5Dget_type() is a named data
type or a read-only transient data type depending on whether
the dataset points to a named data type. The old behavior,
to return a modifiable transient type, is accomplished by
overloading H5Tcopy() to operate on datasets (see above).
* H5Tshare(), H5Tunshare(), and H5Tis_shared() have been
removed from the API.
The following features were *not* implemented because they
need more discussion:
* A named data type can be opened by applying H5Topen() to a
dataset in which case the data type is the data type of the
dataset (or the data type to which the dataset points if the
dataset has a shared data type).
* A named data type can have attributes like groups or
datasets.
* The members of a compound data type can point to named data
types.
./src/h5ls.c
Reports `Data type' for named data type objects in the file.
|
| |
|
|
| |
Added ./html/h5s.examples.
|
| |
|
|
| |
to ':' to make it a null action.
|
| |
|
|
|
| |
in which access_parms is now a pointer to a struct (was a struct
before.)
|
| |
|
|
|
| |
assert. That allowed more customization in the future (like
allow different levels of verboseness.)
|
| |
|
|
|
| |
the tests pass. Should add a "noclean" option that allows the
temporary to stay around even when the tests pass.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
----------------------
./html/Datasets.html
Fixed a couple of typos.
./src/H5.c
Added the `Z' modifier to HDfprintf() for `size_t' sizes. Use
it like this:
HDfprintf(stderr,"size is %Zd\n", (size_t)x);
./src/H5AC.c
./src/H5F.c
./src/H5Fprivate.h
The maximum number of meta data objects that can be cached can
be set from the application (but the library might not honor
it every time; it's a hint).
./src/H5D.c
Changed a warning message so it's not so alarming.
./src/H5Fistore.c
Chunks can be cached.
./src/H5O.c
./src/H5Oprivate.h
Added H5O_copy() and H5O_free() to copy and free messages.
./src/H5P.c
./src/H5Ppublic.h
Added H5Pset_cache() and H5Pget_cache() and changed lots of
"template" to "property list".
./src/H5Z.c
./src/H5Zpublic.h
Miscellaneous little things to clean up. Mostly just removed
H5Z_MAXVAL and added H5Z_USERDEF_MIN and H5Z_USERDEF_MAX.
./MANIFEST
./test/Makefile.in
./test/chunk.c [NEW]
Added a performance test for chunk caching. It looks at the
amount of I/O instead of timing because timing is partly
dependent on the chunk size and I wanted a measurement that
was a function of only the cache size. Run `chunk' with no
arguments and then say `gnuplot x-gnuplot' to see the plots
(press return between plots). Postscript files are created for
each plot.
./test/big.c
./test/cmpd_dset.c
./test/extend.c
./test/external.c
./test/gheap.c
Added H5F_ACC_DEBUG so we can see cache performance
statistics.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Change in Documentation for the INSTALL process on the ASCI Red
Problem:
As part of the parallel HDF5 install process, upon compiling the
library and sequential testers, it is necessary to copy the testers
onto local disk prior to testing. This is almost done by steps 9-11)
...
9) cp -r ../hdf5 ../scratch
10) cd /scratch/hdf5/test
11) make test >&! test.out
After the copying to /scratch, when the "make test" is run,
dependencies are checked and it is determined that the executables
need to be rebuilt. Unfortunately, the MPI-IO libraries in
sasn100:/usr/community/mpi-io are not accessible on ASCI Red, so
that step #11 above gives an error.
Solution:
Albert Cheng and Robb Matzke suggested to add the "-p" option in
step #9, if available, to maintain the original modifications.
Those install steps were changed to:
...
9) cp -rp ../hdf5 ../scratch
10) cd /scratch/hdf5/test
11) make test >&! test.out
Everything works fine!
Platform tested:
janus - ASCI Red
sasn100 - TFLOPS Application server
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
----------------------
./src/H5Bprivate.h
./src/H5F.c
./src/H5Ffamily.c
./src/H5Fprivate.h
./src/H5Gpkg.h
./src/H5MF.c
./src/H5P.c
./src/H5Ppublic.h
./test/big.c
./html/Big.html
./html/Files.html
Family members can now be any size >1kB. Got rid of some
places where we were reading a property list after it was
closed.
./MANIFEST
./src/Makefile.in
./src/h5repart.c
A program to repartition file families. The source and/or
destination may be files or file families. Examples:
testhdf5 -c -o stab
h5repart -m 1m tstab2.h5 x%05d.h5
h5repart -m 100k x%05d.h5 y%05d.h5
h5repart y%05d.h5 z.h5
diff tstab2.h5 z.h5
./src/h5ls.c
Added a usage message, replaced assertions with error
messages.
./config/linux
Made a warning message more obvious. Added better
optimization flags for Pentium-Pro's.
|
| |
|
|
| |
automagically increments the minor release number.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Bug fix
Problem:
On the ASCI Red machine, the configure step is executed from the platform
cross compiler, so that the configure test for "long long int" format support
was inconclusive.
Solution:
It was necessary to uncomment the line
#hdf5_cv_printf_ll=${hdf5_cv_printf_ll='ll'}
in config/intel-osf1 to force "long long int" format support.
Platform tested:
TFLOPS Application server, Intel TFLOPS
|
| |
|
|
|
|
|
|
| |
----------------------
./src/H5G.c
Fixed a read/write of free memory that my home-brew purify
doesn't catch (yet :-)
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
./src/H5F.c
./src/H5G.c
./src/H5O.c
./src/H5Osdspace.c
./src/H5T.c
./src/H5Tconv.c
./src/H5Z.c
./test/big.c
./test/cmpd_dset.c
./test/dsets.c
./test/extend.c
./test/istore.c
Now that I have a home-grown version of Purify I fixed all the
leaks in all the test files.
|
| | |
|
| | |
|
| |
|
|
| |
attributes.
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
----------------------
./src/H5A.c
./src/H5T.c
Fixed memory leaks. More to come later but PureAtria doesn't
make a Linux version of purify and the free version doesn't
compile with the new SMP Linux kernels so I had to debug over
the internet on a day that Sprint seemed to be having routing
problems... oh well. I got rid of most of the leaks.
./src/H5Apublic.h
Includes H5Ipublic.h for types in the header file.
./src/H5O.c
Comments improved for H5O_read()
./test/tattr.c
Removed a non-ANSI empty initializer.
./test/dsets.c
Include <string.h>
./test/istore.c
Fixed a non-ANSI pointer conversion.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
./configure.in
./config/BlankForm
./src/H5.c
Configure tries to figure out how to print `long long' types
and then defines PRINTF_LL_WIDTH to be `ll' or `q' or
something. It does this by running little sprintf() programs
with various formats until it gets one that works. If this
causes problems when cross compiling (like with parallel
machines) then you may add the following to the appropriate
config file:
hdf5_cv_printf_ll=${hdf5_cv_printf_ll='ll'} or
hdf5_cv_printf_ll=${hdf5_cv_printf_ll='q'} or
hdf5_cv_printf_ll=${hdf5_cv_printf_ll='l'}
./config/intel-osf1
./config/irix5.3
Updated these config files to match the others. This allows
the `--enable-production' configure flag to work properly.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
./src/H5D.c
./src/H5Dprivate.h
./src/H5Dpkg.h [REMOVED]
./src/H5Tpkg.h
./src/H5Tprivate.h
Rearranged some header stuff so H5A doesn't have to include
the package-private header files that don't belong to it.
./src/H5E.c
Added the H5E_ATTR message to the major error struct.
|
| |
|
|
|
|
| |
Added INSTALL_parallel.ascired, examples/Attributes.txt,
html/Attributes.html, testpar/Makefile.ascired. Removed
H5Dpkg.h
|
