Purpose:
New Feature; Optimization; Clean-up
(Merged from 1.4 branch)
Description:
There is no symbolic constant to pass to functions that take an
optional object ID for when the caller wants to indicate no object
ID. In the past the caller always passed a negative integer.
GPFS performs poorly.
The h5ls tool decides whether to list the file name in the output
based on a compile-time choice, which isn't always optimal at run
time.
Solution:
Added a symbolic constant H5I_INVALID_HID.
Added code to tell the mmfsd of GPFS to forego byte range token
prefetching.
h5ls decides whether to print the file name at runtime based on
the number of objects being listed.
Platforms tested:
SuSE Linux (arborea), gcc and mpich-1.2.4
SunOS (baldric), gcc
2002-12-03 22:50:23 Robb Matzke <matzke@arborea.spizella.com>
*: Added H5E_FCNTL minor error number.
Purpose:
New Feature; Optimization; Clean-up
(Merged from 1.4 branch)
Description:
There is no symbolic constant to pass to functions that take an
optional object ID for when the caller wants to indicate no object
ID. In the past the caller always passed a negative integer.
GPFS performs poorly.
The h5ls tool decides whether to list the file name in the output
based on a compile-time choice, which isn't always optimal at run
time.
Solution:
Added a symbolic constant H5I_INVALID_HID.
Added code to tell the mmfsd of GPFS to forego byte range token
prefetching.
h5ls decides whether to print the file name at runtime based on
the number of objects being listed.
Platforms tested:
SuSE Linux (arborea), gcc and mpich-1.2.4
SunOS (baldric), gcc
2002-12-03 23:23:45 Robb Matzke <matzke@arborea.spizella.com>
* H5_trace: Added printing for H5S_sel_type, code "St".
2002-12-03 22:45:11 Robb Matzke <matzke@arborea.spizella.com>
*: Removed a little white space. Other changes are probably due to
additional API tracing statements inserted
automatically.
Purpose:
Feature; Optimization
Description:
Clients pass `-1' or make their own #define for HDF5 functions
that take an optional object ID.
Blue's GPFS is slow for typical SAF restart dumps.
Solution:
Added a #define for H5I_INVALID_HID
Added GPFS-specific code to H5FDmpiposix.c that tells mmfsd to
forego byte range token prefetching. This code can be compiled
into the library by defining USE_GPFS_HINTS. The plan is to either
generalize this so it's detected during configure and turned
on/off at runtime, or to move it up into DSL/SAF with the new HDF5
functions to that return the low-level file handle.
Platforms tested:
SuSE Linux (arborea), gcc and mpich-1.2.4
SunOS (baldric), gcc
2002-09-05 12:24:28 Robb Matzke <matzke@arborea.spizella.com>
*: Added #define for H5I_INVALID_HID. An invalid object handle is
actually any integer that isn't currently in use as a
handle, and non-positive integers are never used as a
handle. The #define is -1, and is mostly so that HDF5
users can pass a symbolic name instead of a negative
number to functions that take optional object ID
arguments.
Purpose:
Feature; Optimization
Description:
Clients pass `-1' or make their own #define for HDF5 functions
that take an optional object ID.
Blue's GPFS is slow for typical SAF restart dumps.
Solution:
Added a #define for H5I_INVALID_HID
Added GPFS-specific code to H5FDmpiposix.c that tells mmfsd to
forego byte range token prefetching. This code can be compiled
into the library by defining USE_GPFS_HINTS. The plan is to either
generalize this so it's detected during configure and turned
on/off at runtime, or to move it up into DSL/SAF with the new HDF5
functions to that return the low-level file handle.
Platforms tested:
SuSE Linux (arborea), gcc and mpich-1.2.4
SunOS (baldric), gcc
Purpose:
New feature to H5Dget_offset
Description:
If user block is set, H5Dget_offset should be able to return the absolute
offset from the beginning of file.
Platforms tested:
eirene, arabica
Update
Description:
Cleaned up the debug printing of the generic property list. It was
printout NULL property values for non-existant properties and putting
multiple properties into one group.
Platforms tested:
Eirene (too small change for full testing)
Code cleanup
Description:
Finish checkin of H5A API cleanups with header files I forgot on the first
checkin pass.
Platforms tested:
Tested h5committest {arabica (fortran), eirene (fortran, C++)
modi4 (parallel, fortran)}
FreeBSD 4.7 (sleipnir)
Code Cleanup & New Feature
Description:
H5config.h.in:
Removed H5_HAVE_COMPRESSION & H5_HAVE_FILTER_GZIP flags.
Added H5_HAVE_FILTER_DEFLATE flag.
H5Z.c:
H5Zprivate.h:
H5Zpublic.h:
Switched from using H5_HAVE_COMPRESSION flag in favor of
H5_HAVE_FILTER_DEFLATE.
Added H5Zunregister & H5Zfilter_avail API functions.
Changed a numeric constant (256) to a symbolic constant
(H5Z_FILTER_RESERVED).
Automatically add the shuffling filter to the list of available filters
(when it is enabled).
Moved prototypes for H5Z_filter_deflate & H5Z_filter_shuffle from the
public header into the private header.
H5Zdeflate.c:
Switched from using H5_HAVE_COMPRESSION & H5_HAVE_FILTER_GZIP flags in
favor of H5_HAVE_FILTER_DEFLATE.
Cleaned up formatting & error reporting a bit.
H5Zshuffle.c:
Rewrote shuffling algorithm to be more efficient.
Added error checking & reporting.
Added standard Pablo information.
Added standard function header comment.
Added FUNC_ENTER & FUNC_LEAVE macros.
Platforms tested:
Tested h5committest {arabica (fortran), eirene (fortran, C++)
modi4 (parallel, fortran)}
FreeBSD 4.7 (sleipnir)
Code cleanup, Feature enhancement
Description:
Improve hashing algorithm to not generate as many clashes.
Avoid using H5MD_realloc when changing the size of a chunk on disk.
Instead, use H5MF_free followed by H5MF_alloc, which avoids copying the
information from the space for the previous chunk to the space for the
new chunk, since we are going to overwrite the chunk with new data anyway.
This change should both improve I/O performance in situations where
compressed chunks are being overwritten as well as reduce the space used
in the file during that situation.
Cleaned up error handling in a few cases.
Removed old debugging printf()s.
Platforms tested:
Tested h5committest {arabica (fortran), eirene (fortran, C++)
modi4 (parallel, fortran)}
FreeBSD 4.7 (sleipnir)
Bug Fix
Description:
Free blocks of space in the file which were merged with a block before them
were not updating the free list information correctly, allowing space in
the file to over-allocated and potentially corrupted.
Solution:
Correct address of free block during merging.
Platforms tested:
Tested h5committest {arabica (fortran), eirene (fortran, C++)
modi4 (parallel, fortran)}
FreeBSD 4.7 (sleipnir)
Purpose:
Adding internal shuffle filter
Description:
With the combination of shuffling filter with general compression
algorithm, the compression ratio may be improved without adding much
encoding and decoding time for many real NASA datasets(especially
floating data) and other application
datasets(See techNotes).
Solution:
SHuffle the bytes within the data to utilize the locality.
Platforms tested:
arabica , eirene, modi4
Misc. update:
Update MANIFEST if you add or remove any file.
Update release_docs/RELEASE for bug fixes, new features, etc.
Update applicable document files too.
FPHDF5 Fixups
Description:
- Stopped using API calls in the H5FPclient code. It now uses
internal library calls.
- Removed some FIXME comments because they've been fixed.
- Small fix for an enum starting at 37 for no reason :-)
Platforms tested:
Eirene (only affects FPHDF5, so no need to test 3 platforms just yet).
Code cleanup
Description:
Added some comments and made some minor code cleanups
Platforms tested:
minor change, only testing on FreeBSD 4.7 (sleipnir) w/parallel
Update
Description:
Changed some of the asserts to check that a non-NULL H5F_t* is passed
in.
Platforms tested:
Eirene PP
Arabica Fortran
Modi4 PP Fortran
Update
Description:
Added code which performs an update when there's a dataset creation.
Commented on some FIXMEs which were in the code.
Platforms tested:
Eirene PP
Arabica Fortran
Modi4 PP Fortran
Code Movement
Description:
Moved definition of the H5D_t data structure into the H5Dprivate.h
header file. With the FPHDF5 stuff, we need access to this data
structure.
Platforms tested:
Eirene...change small, so no need for all three tests.
New feature.
Description:
Add in '--enable-filters=filter1,filter2' flag to configure script. This
can also be used as '--enable-filters=all' or just '--enable-filters' to
enable all the internal I/O filters in the library.
This flag can also be set to '--disable-filters' to disable all the internal
I/O filters. Individual filters can not be disabled due to the limitations
of autoconf.
Additionally, added "H5_HAVE_FILTER_<foo>" macros to H5pubconf.h header
file, so the presence of each filter can be detected.
This is in preparation for Kent's shuffling filter to be imported shortly.
Platforms tested:
Tested h5committest {arabica (fortran), eirene (fortran, C++)
modi4 (parallel, fortran)}
Tested FreeBSD 4.7 (sleipnir) also
Fix Windows portability bug
Description:
"long long" is not defined on Windows.
Solution:
Use our internal definition of "long_long" instead.
Platforms tested:
eyeballed.
Fix Windows portability bug
Description:
"long long" is not defined on Windows.
Solution:
Use our internal definition of "long_long" instead.
Platforms tested:
eyeballed.
Code Cleanup
Description:
A function was returning "NULL" instead of "FAIL" for an error
condition. Renamed the variable "class" to "h5_class" since class is
a keyword in C++ and might mess up with a C++ compiler...
Platforms tested:
Eirene pp (too small a change for 3 tests)
Code cleanup
Description:
Cleaned up some error values that were incorrect, added some assertions
and refactored the integer & floating-point native conversions some.
Platforms tested:
FreeBSD 4.7 (sleipnir), too small for h5committest
Purpose:
New function.
Description:
H5Dget_offset returns the offset of a dataset's data relative to the
beginning of a file.
Platforms tested:
arabica(simple function, one test should be enough.)
Misc. update:
RELEASE.txt
New Feature. Support for new FPHDF5 Feature
Description:
More support for the FPHDF5 feature. This splits the H5D_create
function into separate parts which update the metadata cache. It was
necessary to split apart the H5O_create function to do a similar
thing since it would allocate real space on the hard disk.
I'm checking this up now so that I don't get too far away from what
the CVS repository has and so that others may look at the code and
retch^H^H^H^H^Hmarvel at it. If there are any comments, send them my
way.
There is one outstanding issue, though. I need to use non-API
functions in the FPHDF5 stuff. I'm using some APIs right now and
should migrate to using others. At the moment, I'm putting that on
the back burner until I'm more along in the implementation.
Platforms tested:
Arabica (C++)
Eirene (Parallel)
Modi4 (Parallel)
New feature
Description:
Add support for scalar dataspaces in parallel I/O.
Platforms tested:
Tested h5committest {arabica (fortran), eirene (fortran, C++)
modi4 (parallel, fortran)}
Also, FreeBSD 4.7 (sleipnir) serial & parallel
Misc. update:
Update release_docs/RELEASE for bug fixes, new features, etc.
Bug fix & code cleanup
Description:
Hyperslab code for collapsing dimensions was incorrectly collapsing
selections inappropriately when the fastest changing dimension couldn't
be collapsed.
Also add some more assertions which will make similar bugs easier to find.
Solution:
Break out of loop earlier.
Platforms tested:
modi4 (parallel), too small to triple check.
Code cleanup
Description:
Add more comments and clean up small bits of the FPH5 code.
Platforms tested:
FreeBSD 4.7 (sleipnir), changes too minor to affect other platforms.
Another Oops
Description:
I accidentally checked in the H5D.c module when I shouldn't have.
Solution:
Rolled back to 1.281 version
Platforms tested:
Linux
Oops
Description:
The H5FPprivate.h file was being #included even if PARALLEL was
turned off.
Solution:
put the #include of the H5FPprivate.h file within the #ifdef block so
that it doesn't get included if H5_HAVE_FPHDF5 isn't defined.
Platforms tested:
Linux...
Bug Fix
Description:
The NBYTE{ENCODE,DECODE} macros were updating the pointer "p" to a
new value instead of one of the "real" pointers passed into the
macro.
Solution:
Changed it so that it updates the destination pointer which gets
passed into these macros.
I'm not sure if this is correct, since neither of these macros are
actually used in our source code...
Platforms tested:
They aren't used, so no testing possible.
Feature Add
Description:
New files for the Flexible Parallel HDF5 stuff.
H5FP.c - Module housing the APIs to FPHDF5
H5FPclient.c - Module housing the internal client APIs
H5FPserver.c - Module housing the internal server APIs
H5FPpublic.h - Header for public APIs
H5FPprivate.h - Header for private APIs
H5Ofphdf5.c - Way of serializing FPHDF5 information to and from the
SAP
H5Oplist.c - Way of serializing a generic property list.
Solution:
[details about the changes, algorithm, etc...]
[Please as detail as you can since your own explanation is
better than others guessing it from the code.]
Platforms tested:
Tested h5committest {arabica (fortran), eirene (fortran, C++)
modi4 (parallel, fortran)}?
[If no, why not?]
Other platforms/configurations tested?
Misc. update:
Update MANIFEST if you add or remove any file.
Update release_docs/RELEASE for bug fixes, new features, etc.
Update applicable document files too.
