Lots of performance improvements & a couple new internal API interfaces.
Description:
Performance Improvements:
- Cached file offset & length sizes in shared file struct, to avoid
constantly looking them up in the FCPL.
- Generic property improvements:
- Added "revision" number to generic property classes to speed
up comparisons.
- Changed method of storing properties from using a hash-table
to the TBBT routines in the library.
- Share the propery names between classes and the lists derived
from them.
- Removed redundant 'def_value' buffer from each property.
- Switching code to use a "copy on write" strategy for
properties in each list, where the properties in each list
are shared with the properties in the class, until a
property's value is changed in a list.
- Fixed error in layout code which was allocating too many buffers.
- Redefined public macros of the form (H5open()/H5check, <variable>)
internally to only be (<variable>), avoiding innumerable useless
calls to H5open() and H5check_version().
- Reuse already zeroed buffers in H5F_contig_fill instead of
constantly re-zeroing them.
- Don't write fill values if writing entire dataset.
- Use gettimeofday() system call instead of time() system when
checking the modification time of a dataset.
- Added reference counted string API and use it for tracking the
names of objects opening in a file (for the ID->name code).
- Removed redundant H5P_get() calls in B-tree routines.
- Redefine H5T datatype macros internally to the library, to avoid
calling H5check redundantly.
- Keep dataspace information for dataset locally instead of reading
from disk each time. Added new module to track open objects
in a file, to allow this (which will be useful eventually for
some FPH5 metadata caching issues).
- Remove H5AC_find macro which was inlining metadata cache lookups,
and call function instead.
- Remove redundant memset() calls from H5G_namei() routine.
- Remove redundant checking of object type when locating objects
in metadata cache and rely on the address only.
- Create default dataset object to use when default dataset creation
property list is used to create datasets, bypassing querying
for all the property list values.
- Use default I/O vector size when performing raw data with the
default dataset transfer property list, instead of querying for
I/O vector size.
- Remove H5P_DEFAULT internally to the library, replacing it with
more specific default property list based on the type of
property list needed.
- Remove redundant memset() calls in object header message (H5O*)
routines.
- Remove redunant memset() calls in data I/O routines.
- Split free-list allocation routines into malloc() and calloc()-
like routines, instead of one combined routine.
- Remove lots of indirection in H5O*() routines.
- Simplify metadata cache entry comparison routine (used when
flushing entire cache out).
- Only enable metadata cache statistics when H5AC_DEBUG is turned
on, instead of always tracking them.
- Simplify address comparison macro (H5F_addr_eq).
- Remove redundant metadata cache entry protections during dataset
creation by protecting the object header once and making all
the modifications necessary for the dataset creation before
unprotecting it.
- Reduce # of "number of element in extent" computations performed
by computing and storing the value during dataspace creation.
- Simplify checking for group location's file information, when file
has not been involving in file-mounting operations.
- Use binary encoding for modification time, instead of ASCII.
- Hoist H5HL_peek calls (to get information in a local heap)
out of loops in many group routine.
- Use static variable for iterators of selections, instead of
dynamically allocation them each time.
- Lookup & insert new entries in one step, avoiding traversing
group's B-tree twice.
- Fixed memory leak in H5Gget_objname_idx() routine (tangential to
performance improvements, but fixed along the way).
- Use free-list for reference counted strings.
- Don't bother copying object names into cached group entries,
since they are re-created when an object is opened.
The benchmark I used to measure these results created several thousand
small (2K) datasets in a file and wrote out the data for them. This is
Elena's "regular.c" benchmark.
These changes resulted in approximately ~4.3x speedup of the
development branch when compared to the previous code in the
development branch and ~1.4x speedup compared to the release
branch.
Additionally, these changes reduce the total memory used (code and
data) by the development branch by ~800KB, bringing the development
branch back into the same ballpark as the release branch.
I'll send out a more detailed description of the benchmark results
as a followup note.
New internal API routines:
Added "reference counted strings" API for tracking strings that get
used by multiple owners without duplicating the strings.
Added "ternary search tree" API for text->object mappings.
Platforms tested:
Tested h5committest {arabica (fortran), eirene (fortran, C++)
modi4 (parallel, fortran)}
Other platforms/configurations tested?
FreeBSD 4.7 (sleipnir) serial & parallel
Solaris 2.6 (baldric) serial
fortran/config/powerpc-ibm-aix5.x
To cover AIX 5.x.
Removed Files:
fortran/config/powerpc-ibm-aix4.2.1.0
fortran/config/powerpc-ibm-aix4.3.2.0
These files are outdated and have been replaced by powerpc-ibm-aix4.x.
Modified Files:
MANIFEST
Update it.
Platforms tested:
Only on LLNL blue. LLNL SP are the only machines that use
these files and that I have access.
Update
Description:
Added new files to the MANIFEST and added a new configure switch for
enabling building of the Flexible parallel HDF5 module.
Platforms tested:
Modi4 Eirene Arabica
Purpose:
Bug fix, code improvement
Description:
Fortran tests didn't cleanup created files. Also HDF5_PREFIX and
HDF5_PARAPREFIX were not used to specify location of the files.
There was a redundant file in the testpar directory that contained
an error reporting function used by both serial and parallel tests.
Solution:
Created library h5test_fortran.a that contains functions used by the
serial and parallel tests. It includes Fortran and C functions that
may be called from Fortran programs to report errors, to modify
file names and to cleanup files after run. Modified test code to
use new functions.
Platforms tested:
Linux 2.2 (eirene) serial
IRIX64 (modi4) parallel with HDF5_PREFIX and HDF5_PARAPREFIX set to $SCR
Solris 2.7 with mpich 1.2.4 with HDF5_PARAPREFIX set to /tmp/epourmal
Purpose:
Design for compact dataset
Description:
Compact dataset is stored in the header message for dataset layout.
Platforms tested:
arabica, eirene.
Code cleanup
Description:
Move get/set routines for each type of property list (file creation,
dataset creation, file access and dataset transfer) into their own source
code module.
Platforms tested:
FreeBSD 4.6 (sleipnir)
Compress I/O Test Add
Description:
This is the initial check-in of the compress I/O test. It will write
out compressed buffers to a file.
I need to implement a write of uncompressed data first for
comparison...
Platforms tested:
Linux
Update
Description:
Updated the way we used the AC_INIT macros so that it's current to
the new standard way and not deprecated anymore.
Also, added a test for some functions during parallel Fortran
configure. This required the creation of an H5config_fortran.h file
so that we can test for these...
Platforms tested:
Elena is going to test the changes to the fortran after she checks in
her other code. I will test on SDSC machines after getting her code.
The other changes were tested on Linux.
Autotools Update
Description:
I've updated autoconf, automake, and libtool to the latest/greatest
versions; 2.53, 1.6, and 1.4.2 resp.
Many changes come with the new versions:
- ltconfig is no longer used
- acconfig.h is no longer used (#define values are declared with
the macro)
- regeneration of all of the aclocal.m4, configure, and
H5config.h.in files.
- new config.{guess,sub} files
- new ltmain.sh file
Platforms tested:
AIX (blue), and Linux
Purpose:
Feature Add
Description:
Added support for dumping Group Comments. This involved a
modification of the DDL as well.
Solution:
Steal code from h5ls and put it in the h5dump. The ddl.html file was
updated as normal. And a test was created...
Platforms tested:
Dangermouse, Modi4, Kelgia
Removing the DPSS (gridstorage) driver source code.
Description:
The DPSS (using Grid-Storage) driver is retired.
Removed the configure option with-gridstorage from configure.in.
Cvs remove the following files
./src/H5FDdpss.c
./src/H5FDdpss.h
./test/dpss_read.c
./test/dpss_write.c
Regenerated Dependencies files (some had to be hand-edited since
'make depend' did not cover them.)
Removed reference to DPSS Virtual file driver from H5F.c.
Platforms tested:
modi4 (Parallel; -with-gass=...), eirene, arabica (fortran, cxx).
Purpose:
Feature Add
Description:
Added "install-example" and "install-all" to the Makefile system.
The behaviour of the "make install*" options:
make install - Installs binaries, libraries, include
files, and example programs.
make install-examples - Installs only the example programs.
The directories are:
${prefix}/doc/hdf5/examples/{c,c++,fortran}
make install-all - Install the binaries, libraries, include
files, example programs, and
documentation. The whole kit-n'-caboodle.
make uninstall-examples - Get rid of those example files (but not
the ${prefix}/doc/hdf5/examples/...
directories)
There's a new bin/ program which helps create directories which are
deeply nested called "mkdirs". It's a simple shell script.
Platforms tested:
Linux
New feature
Description:
Added perform programs to test the HDF5 library performance. Programs
are installed in directory perform/.
Platforms tested:
eirene
Purpose:
Bug Fix
Description:
The way we were generating Dependencies and .depend files was broken.
If the $srcdir or other macros began with a ".", then it would match
anything and cause problems since it would then overwrite the
beginning of the header file's path.
Solution:
Wrote a Perl script which can handle this type of weirdness better.
It's only used when the environment is a GNU one with a GCC
compiler...
Platforms tested:
Linux
Purpose:
Maintenance
Description:
Source directory has been rearranged. INSTALL*, HISTORY.txt and RELEASE.txt were moved to the release_docs directory. *.zip files were moved to the windows directory. README file was renamed to README.txt MANIFEST was updated to reflect those changes.
Purpose:
Feature
Description:
Changed Pablo source code to allow the user to build the instrumented
library to link with either the Trace Libraries or Pablo PCF.
Solution:
Added conditionally compiled code to routines in the pablo subdirectory.
Also added a new file PabloHDF_MPIO.c containing all of the MPI I/O code
for better code structure.
Platforms tested:
Solaris 32-bit, Irix 64
Purpose:
Adding two new files (doc/html/fortran/F90Flags.html, and
doc/html/fortran/h5_FORTRAN.html) to FORTRAN portion of
reference manual.
Platforms tested:
IE 5
Purpose:
Bug fix
Description:
There was a typo in the gif.h path that I overlooked first time.
Solution:
Fixed a typo. bin/chkmanifest does not complain anymore.
Platforms tested:
baldric
Purpose:
Adding new feature
Description:
Added gif2h5 and h52gif conversion utilities
Solution:
The utilites follow the framework built for the gif2hdf and hdf2gif
utilities for hdf4. The main files modified were those that read the
H5 file and those that write H5 file. In the future, if you wish to
continue with the framework and extend it to .png or some other fileformat
the main files to edit will be the gif reader and writer.
One point to note with h52gif. You have to specify the exact location of
the image and the palette that it links to. You can choose not to specify
a palette (uniform grayscale chosen in this case) but you must specify
image location. In the future, someone could edit the readhdf.c source
to enable the reader to parse the hdf file and select all images with
corresponding palettes.
Platforms tested:
modi4 , eirene , hawkwind , arabica , Ren (NT 4.0) , Personal box (win2k)
Purpose:
Windows F90 port
Description:
all_withf90.zip file conyains all necessary projects to build
and test HDF5 library (both C and F90) on Windows platforms
Platforms tested:
Windows 98
