Maintenance on Windows
Description:
Move Win32 Macros(snprintf and vsnprintf) to the right places
Solution:
Platforms tested:
MSVS 6.0 on Windows XP
heping
Misc. update:
bug fix
Description:
The set of h5diffdebugN macros that printed nothing did not have
the proper number of parameters. Some compilers (gcc) did not
like it.
Solution:
Fixed them to have the matching number of parameters.
Platforms tested:
heping pp (using both mpich-icc and mpich-gcc), modi4 pp.
Code cleanup
Description:
Clean up code somewhat:
- Move from HDmemset() -> H5G_ent_reset() to clear out group entry info
- Simplify H5G_unlink() call
- Use portability macros instead of direct system calls in more places
- Improve readbility by neatening whitespace, etc.
- Move some macros into source code module instead of header files
Platforms tested:
FreeBSD 4.11 (sleipnir)
Linux 2.4
New debugging feature
Description:
Add some code to track where memory from the internal free list allocator
is allocated within the library. It's not quite ready for "prime time" yet,
but it's close enough to be useful.
This is disabled by default and requires the H5FL_TRACK macro in
src/H5FLprivate.h to be uncommented to activate during debugging.
Platforms tested:
FreeBSD 4.11 (sleipnir)
Linux 2.4
Too minor to require full h5committest
Code cleanup
Description:
Clean up unused variables & use portability macros better
Platforms tested:
FreeBSD 4.11 (sleipnir)
Too minor to require h5committest
Bug fix
Description:
Fix core dump when checking whether to invalidate the name of an object
when unlinking an object in a group opened through an object reference.
Solution:
Check if names of various objects are valid before comparing them, etc.
Platforms tested:
FreeBSD 4.11 (sleipnir)
Too minor to require h5committest
Code cleanup/bug fix
Description:
Add missing files to manifest
Remove C++ style single-line comment from recent checkin and also re-indent
in a few places.
Platforms tested:
FreeBSD 4.11 (sleipnir) w/thread-safe
Not configured w/h5committest
Description:
Forgot to modify the examples for the "using" declaration changes
in the library. Added "using" declaration appropriately.
Platforms tested:
Linux 2.4 (heping) - very minor
Improvement on windows
Description:
Add Thread safe project file with pthread-win32 on Windows
Solution:
Platforms tested:
MSVS 6.0 on Windows XP
Misc. update:
Improvement
Description:
ttsafe cancel failed on Windows because pthread mutex and cond are not initialized
Add some assertion statements to locate the errors when they happen
Solution:
Initialize pthread mutex and cond to the default attributes.
Platforms tested:
MSVS 6.0 on windows XP and heping(with pthread)
Misc. update:
Improvement
Description:
Add some assertion statements to locate the errors when they happen
Solution:
Platforms tested:
MSVS 6.0 on windows and heping(with pthread)
Misc. update:
Improvement
Description:
Add some assertion statements to locate the error when errors happen
Solution:
Platforms tested:
MSVS 6.0 on windows XP and heping(enable threadsafe with pthread)
Misc. update:
Add more comments for collective chunk IO supports.
Description:
On some platforms that use mpich library internally, we have no way to check mpich version number. We have to manually turn off the collective IO supports.
One example is Altix(cobalt). Just put a comment there so that we won't forget about this.
Solution:
Platforms tested:
Misc. update:
New feature
Description:
All datasets were 1D only before. Added -g to use 2D datasets such
that it allows more options in testing various access patterns.
Platforms tested:
heping pp.
Misc. update:
Feature and bug patch
Description:
Added the SKIP feature which skips a test.
The latest code will hang the very last test. Skip that test for now.
Platforms tested:
heping pp.
Description: repart_test.c was mistakenly treated as a tool, instead of a test program.
Solution: Took it out of the tool list.
Platforms tested: fuss - simple change.
Bug fix
Description:
ph5diff had been hanging in Tflops. Found out that vsnprintf ph5diff
uses was a local coded that did not limit printing according to the size
argument. That resulted in buffer overflow and other problems.
Solution:
Added some sort of size checking in the home-grown vsnprintf and
had ph5diff checked for error return of vsnprintf. Leon also
revamped the ph5diff manager's way of handling communications with
the workers. That eliminated all but the last case of hanging.
Platforms tested:
Tflops.
Misc. update:
Bug fix.
Description:
Reactivated the code for snprintf and vsnprintf for Tflops but added
the SN_SIZ_MIN as an attempt to require the minimum amount of space needed,
hoping vsnprintf/snprintf do not print larger than it per request.
Also added code to detect overflow and return error (IF overflow has
not crashed the application yet.) These are all short term patches.
Platforms tested:
Tested at Tflops.
Update & feature
Added a new cached data conversion setting.
Added -DH5_FLOPS to CPPFLAGS for compiling. Had used the builtin
__PUMAGON__ but H5_TFLOPS is more meaningful.
Platforms tested:
Tested in Flops.
Misc. update:
Description: In some macro definition, some constant values like D_MIN and D_MAX were
not casted to the destination type. They may cause problems for some system like
TFLOPS.
Solution: Cast them to destination type.
Platforms tested: fuss - very simple changes.
Maintenance on windows
Description:
Add two macros for windows, susbstitue functions HDsnprintf and HDvsnprintf
with _snprintf and _vsnprintf when using Visual C++ on windows
Solution:
Platforms tested:
MSVC 6.0 on Windows XP
heping
Misc. update:
Code cleanup
Description:
Trim trailing whitespace, which is making 'diff'ing the two branches
difficult.
Solution:
Ran this script in each directory:
foreach f (*.[ch] *.cpp)
sed 's/[[:blank:]]*$//' $f > sed.out && mv sed.out $f
end
Platforms tested:
FreeBSD 4.11 (sleipnir)
Too minor to require h5committest
Code cleanup
Description:
Collapse the various tests on groups opened by dereferencing into one test.
Platforms tested:
FreeBSD 4.11 (sleipnir)
Too minor to require h5committest
Code cleanup
Description:
Use enum type instead of int for datatype conversion exception callback
routine.
Platforms tested:
FreeBSD 4.11 (sleipnir)
Too minor to require h5committest
Code cleanup
Description:
Fix logic error in previous checkin and also finish refactoring I/O
initialization, including simplifying all the collective & parallel cases
into a more unified mechanism.
Platforms tested:
FreeBSD 4.11 (sleipnir) w/ & w/o parallel
Linux 2.4 (mir)
Bug fix
Description:
Fix logic error for non-parallel file I/O with library built for parallel
I/O capability.
Platforms tested:
Linux 2.4 (mir) w/parallel
New Feature.
Description:
Added code and changes so that h5perf (pio) can be built by
standalone mode (that is built by just "h5pcc" without all these
internal files.) With the standalone mode, h5perf can be built
against different versions of hdf5 library and to contrast the
performance among them.
(Note that pio_standalone.c and pio_standalone.h are used in
standalone build only.)
Platforms tested:
Tested in copper and heping.
Code cleanup & bug fix
Description:
Refactor recent collective chunk changes to simplify them. Also, compute
the number of chunks by checking the number of objects in the chunk skip list
instead of getting the boundaries of the selection and counting the chunks that
it overlaps, which is faster and computes correct number for certain selections.
Platforms tested:
FreeBSD 4.11 (sleipnir) w/parallel
Too minor to require h5committest
Code cleanup
Description:
Refactor, simplify and cleanup Kent's recent checking for collective
chunk I/O. There's a bug that I need to talk to Kent about and some more
cleanups still, but this is reasonable for an interim point.
Platforms tested:
FreeBSD 4.11 (sleipnir) w/parallel
Too minor for h5committest
bug fix.
Description:
The code for functions snprintf and vsnprintf ignored the size parameter.
It could overflow the buffer. The two functions are supported by many
platforms now. Disable the code and see what platform(s) still need
it.
Platforms tested:
h5committested
Misc. update:
Description: In the macro H5T_CONV_Xx_CORE definition, one line said
"} else if (*((ST*)S) < (D_MIN)) {". It caused problem on TFLOPS because
it defines LLONG_MIN as a hexadecimal value instead of decimal value.
Solution: Changed the line to "} else if (*((ST*)S) < (DT)(D_MIN)) {" to
avoid the problem. (Need to change other similar places when time permits.)
Platforms tested: fuss - simple change.
Modified collective chunk Io test
Description:
One test(collective chunk IO test 4) is no longer needed. Comment out first.
The code should be cleaned up later.
Solution:
Platforms tested:
Misc. update:
bug fix for collective chunk IO, phase 1
Optimization hasn't been done yet, the collective chunk IO bug should be fixed.
Description:
In chunking storage, memory space and file space will be remapped, So to check
whether file space and memory space are regular in order to use optimized MPI derived
datatype for collective call one has to check per-chunk wise instead of per hyperslab wise.
Even a regular memory space will be stored in span-tree and will be irregular before chunk IO.
Solution:
1. Check file space and memory space per chunk wise instead of per hyperslab wise.
2. For collective IO mode, number of chunks covered by hyperslab may be different. Since we are
handing per chunk per IO, for the extra chunk IO for some(not all) processors, collective mode will
cause program hanged. So for the extra chunk Io mode independent IO has to be used.
3. On some platforms, Complex MPI derived datatype is not working, so we have to use independent IO for collective IO mode if the selection is irregular. However, when the selection is regular, we do want to use collective IO since that will improve performance. Special cares have to be added for this case.
Platforms tested:
copper(AIX 5.1) Linux(heping mpich 1.2.6), Teragrid machine, Cobalt(altix), modi4
Misc. update:
Big fix
Description:
A group opened by dereferencing a object reference would not work for
H5Giterate(), due to the local heap & B-tree information not being cached.
Solution:
Get the local heap & B-tree info & point to that structure instead of
the group entry for the group.
Platforms tested:
FreeBSD 4.11 (sleipnir)
Too minor to require h5committest
Code cleanup
Description:
Add a check for the correct type of ID for H5Eclose_stack()
Platforms tested:
FreeBSD 4.11 (sleipnir)
Too minor to require h5committest
Description: The change made yesterday regarding repart_test.c wasn't right.
Solution: Roll back to v1.6 until a solution is figured out.
Platforms tested: fuss - simple change.
