Bug fix/optimization
Description:
Address slowdown in MPI-I/O file metadata operations that was introduced
mid-stream. We now _require_ a POSIX compliant parallel file system for the
MPI-I/O file driver (as well as for the MPI-POSIX file driver).
Also optimized file open operation when the file is being created by
reducing the number of collective & syncronizing calls.
Additionally, refactor the MPI routines into a common place, eliminating
duplicated code.
Platforms tested:
FreeBSD 4.9 (sleipnir) w/parallel
h5committest
Bug fix
Description:
Our previous "optimization" of metadata writing which only wrote metadata
from one process was abusing MPI-I/O and after some consultation with Rob Ross
and Rajeev Thakur, Albert & I have come up with a solution...
Solution:
Instead of only writing from one process, issue a collective write
operation with all processes, for metadata writes.
Platforms tested:
FreeBSD 4.9 (sleipnir)
h5committest
Bug fix
Description:
When datasets are deleted from a file, they are removed from the sieve
buffer, but instead of invalidating only the part of the sieve buffer affected,
the sieve buffer code would throw away the entire sieve buffer, potentially
including other raw data in the buffer that hadn't been written to disk yet.
Solution:
Improve the sieve buffer clearing code to handle partial invalidations.
Platforms tested:
FreeBSD 4.8 (sleipnir)
h5committest
Code cleanup
Description:
Various cleanups resulting from running lint tool over H5F.c source module
Platforms tested:
FreeBSD 4.8 (sleipnir)
too minor to require h5committest
Bug fix
Description:
When a non-default indexed storage B-tree internal 'K' value is set by the
user, the chunked datasets created in that file (until it is closed) use the
user's 'K' value and the data can be accessed correctly, but the 'K' value is
not stored in the file.
However, once the file is closed and re-opened, the non-default 'K' value
is lost and the data in the chunked datasets will not be able to be accessed
correctly.
Solution:
Store the indexed storage B-tree internal 'K' value in the superblock.
Platforms tested:
FreeBSD 4.8 (sleipnir)
h5committest
Code cleanup
Description:
Finish converting the B-tree 'K' values to use unsigned integers, rather
than signed ones, since negative amounts of entries in a B-tree node aren't
meaningful.
Platforms tested:
FreeBSD 4.8 (sleipnir)
h5committest
Code cleanup
Description:
Break some of the "debugging" routines into their own module, so they
aren't pulled into every executable, which certainly isn't going to use them.
Platforms tested:
h5committested
Code cleanup
Description:
Remove "UNUSED" keyword from function prototypes in header files.
Platforms tested:
IA64 Linux cluster (titan)
too small to need commit test
Code cleanup/Feature removal
Description:
Remove "round robin" writing of metadata in MPI VFDs (and a couple of other
places). This never really worked the way I hoped it would and it causes
problems on certain machines.
Platforms tested:
FreeBSD 4.8 (sleipnir) w/parallel
Too small for triple check.
Code cleanup
Description:
Limit the scope on more function prototypes/macros/typedefs.
Platforms tested:
FreeBSD 4.8 (sleipnir)
h5committest not necessary.
Code cleanup
Description:
Clean up warnings exposed by compiling on O2K. Also, revert some of Bill
and my changes to the H5S_mpi_opt_types_g, etc. and settle them back into their
original location.
Platforms tested:
h5committested.
New feature/enhancement
Description:
Chunked datasets are handled poorly in several circumstances involving
certain selections and chunks that are too large for the chunk cache and/or
chunks with filters, causing the chunk to be read from disk multiple times.
Solution:
Rearrange raw data I/O infrastructure to handle chunked datasets in a much
more friendly way by creating a selection in memory and on disk for each chunk
in a chunked dataset and performing all of the I/O on that chunk at one time.
There are still some scalability (the current code attempts to
create a selection for all the chunks in the dataset, instead of just the
chunks that are accessed, requiring portions of the istore.c and fillval.c
tests to be commented out) and performance issues, but checking this in will
allow the changes to be tested by a much wider audience while I address the
remaining issues.
Platforms tested:
h5committested, FreeBSD 4.8 (sleipnir) serial & parallel, Linux 2.4 (eirene)
Finish code cleanup
Description:
Wrap up the conversion of H5F_flush's multiple boolean flags into a single
bitfield of flags by pushing the flags down into the H5AC_flush and
H5F_istore_flush routines.
Also, changed the flags from H5_FLUSH_<foo> to H5F_FLUSH_<foo> to be more
consistent with rest of library.
And reverted the changes to H5FDflush and H5FD_flush routines.
Platforms tested:
FreeBSD 4.7 (sleipnir)
Solaris 5.8 (sol)
IRIX64 6.5 (modi4) w/parallel
Misc. update:
Bug Fix
Description:
Metadata cache in parallel I/O can cause hangs in applications which
perform independent I/O on chunked datasets, because the metadata cache
can attempt to flush out dirty metadata from only a single process, instead
of collectively from all processes.
Solution:
Pass a dataset transfer property list down from every API function which
could possibly trigger metadata I/O.
Then, split the metadata cache into two sets of entries to allow dirty
metadata to be set aside when a hash table collision occurs during
independent I/O.
Platforms tested:
Tested h5committest {arabica (fortran), eirene (fortran, C++)
modi4 (parallel, fortran)}
FreeBSD 4.7 (sleipnir) serial & parallel
Misc. update:
Updated release_docs/RELEASE
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
Purpose:
__DLL__ is a keyword in some platforms and __DLL__ is also defined as a macro for windows DLL applications.
That causes problems.
Description:
Solution:
Use H5_DLL*** to replace __DLL***__ at all header files.
Change the macro defination at H5api_adpt.h.
Platforms tested:
linux2.2.18smp, irix64, solaris 2.7 and windows 2000
Bug fix/Code cleanup/New Feature
Description:
Correct problems with writing fill-values to external storage and allocate
the data storage at the correct times.
Also, mostly straighten out the strange code which allocates and fills
raw data storage for datasets. Things are still a bit odd in that the
fill-values for chunked datasets are written when the space is allocated,
instead of in a separate routine, but there are two reasons for this:
it's inefficient (especially in parallel) to iterate through all the chunks
twice, and (more importantly) the space needed to store compressed chunks
isn't known until we've got a buffer of compressed fill-values ready to
write to the chunk.
Additionally, add in the H5D_SPACE_ALLOC_INCR and H5D_SPACE_ALLOC_DEFAULT
setting for the "space time", which incorporate the previous behavior of
the space allocation for chunked datasets.
The default settings for the different types of dataset storage are now
as follows:
Contiguous - Late
Chunked - Incremental
Compact - Early
This checkin also incorporates a change to the behavior of external data
storage in two ways - fill-values are _never_ written to external storage
(under the assumption that writing fill-values is triggered by allocating
space in an HDF5 file, and since space is not allocated in the file, the
fill-values should not be written) and external data files are now created
if they don't exist when data is written to them. The fill-value will
probably need to be revisited at some time in the future, this just seemed
like the safer course currently.
I think I cleaned up some compiler errors also, before getting bogged down
in the fixes for the space allocation and fill-values.
Platforms tested:
FreeBSD 4.6 (sleipnir) w/serial & parallel. Will be testing on IRIX64
6.5 (modi4) in serial & parallel shortly.
Purpose:
Design for compact dataset
Description:
Compact dataset is stored in the header message for dataset layout.
Platforms tested:
arabica, eirene.
New feature.
Description:
Added MPI-posix VFL driver. This driver uses MPI to coordinate actions, but
performs I/O directly with posix sec(2) I/O functions. This driver should
_NOT_ be used if the file accessed is not on a parallel filesystem.
Platforms tested:
FreeBSD 4.6 (sleipnir) w/parallel & IRIX64 6.5 (modi4) w/parallel
Code improvement.
Description:
Added boot block and driver info block checksumming feature to the shared
file information. This prevents these blocks from being written out
multiple times when they haven't changed.
This reduces the number of I/O operations which hit the disk for my test
program from 15 to 14 (i.e. from 393 to 14, overall).
Platforms tested:
Solaris 2.7 (arabica) w/FORTRAN and FreeBSD 4.5 (sleipnir) w/C++
Bug fix/Code improvement.
Description:
Currently, the chunk data allocation routine invoked to allocate space for
the entire dataset is inefficient. It writes out each chunk in the dataset,
whether it is already allocated or not. Additionally, this happens not
only when it is created, but also anytime it is opened for writing, or the
dataset is extended. Worse, there's too much parallel I/O syncronization,
which slows things down even more.
Solution:
Only attempt to write out chunks that don't already exist. Additionally,
share the I/O writing between all the nodes, instead of writing everything
with process 0. Then, only block with MPI_Barrier if chunks were actually
created.
Platforms tested:
IRIX64 6.5 (modi4)
Migrate from configure macros of XYZ_ABC to H5_XYZ_ABC
Description:
configure generates many macros definitions on the fly and
were stored in src/H5config.h which is included by H5public.h.
But other software that uses hdf5 may also run their own configure.
There can be a clash in macro name space. We decided awhile ago
to prepend all generated macros with "H5_" to avoid conflicts.
The process has started and this commit completes it (at least attempt
to).
Solution:
Many macros symbols (e.g. SIZEOF_xxx and HAVE_xxx were changed to
H5_SIZEOF_xxx and H5_HAVE_xxx). Then H5private.h no longer includes
H5config.h. This cuts H5config.h away from HDF5 source code.
Pending issues:
The module of fortran and pablo are to be resolved in a different
commit.
Platforms tested:
eirene (parallel), arabica (solaris 7 --enable-fortran, --enable-cxx)
Code cleanup
Description:
Previously, the I/O pipeline (pline), external file list (efl) and fill-
value (fill) structs were passed down the raw data function call chain,
even into and/or through functions which didn't use them. Since all three
of these pieces of information are available from the dataset creation
property list, just pass the dataset creation property list down the
function call chain and query for the information needed in a particular
function.
Platforms tested:
FreeBSD 4.5 (sleipnir)
Bug Fix & Feature
Description:
The selection offset was being ignored for optimized hyperslab selection
I/O operations.
Additionally, I've found that the restrictions on optimized selection
I/O operations were too strict and found a way to allow more hyperslabs
to use the optimized I/O routines.
Solution:
Incorporate the selection offset into the selection location when performing
optimized I/O operations.
Allow optimized I/O on any single hyperslab selection and also allow
hyperslab operations on chunked datasets.
Platforms tested:
FreeBSD 4.5 (sleipnir)
Purpose:
Modify H5Fclose behavior
Description:
The HDF5 actual file close behaves in several ways in terms of if there
are still objects(dataset, group, datatype) opened in file.
Solution:
Added a new file access property, file close degree. It has four values,
H5F_CLOSE_DEFAULT
H5F_CLOSE_WEAK
H5F_CLOSE_SEMI
H5F_CLOSE_STRONG
The way a file is closed is decided by these values.
Platforms tested:
IRIX64 6.5, SunOS 5.6, FreeBSD 4.4
Code cleanup
Description:
Windows is generating hundreds of warnings from some of the practices in
the library. Mostly, they are because size_t is 32-bit and hsize_t is
64-bit on Windows and we were carelessly casting the larger values down to
the smaller ones without checking for overflow.
Also, some other small code cleanups,etc.
Solution:
Re-worked some algorithms to eliminate the casts and also added more
overflow checking for assignments and function parameters which needed
casts.
Kent did most of the work, I just went over his changes and fit them into
the the library code a bit better.
Platforms tested:
FreeBSD 4.4 (hawkwind)
Code cleanup (sorta)
Description:
When the first versions of the HDF5 library were designed, I remembered
vividly the difficulties of porting code from a 32-bit platform to a 16-bit
platform and asked that people use intn & uintn instead of int & unsigned
int, respectively. However, in hindsight, this was overkill and
unnecessary since we weren't going to be porting the HDF5 library to
16-bit architectures.
Currently, the extra uintn & intn typedefs are causing problems for users
who'd like to include both the HDF5 and HDF4 header files in one source
module (like Kent's h4toh5 library).
Solution:
Changed the uintn & intn's to unsigned and int's respectively.
Platforms tested:
FreeBSD 4.4 (hawkwind)
Bug Fix, Code Cleanup, Code Optimization, etc.
Description:
Fold in the hyperslab speedups, clean up compile warnings and change a
few things from using 'unsigned' or 'hsize_t' to use 'size_t' instead.
Platforms tested:
FreeBSD 4.3 (hawkwind), Solaris 2.7 (arabica), Irix64 6.5 (modi4)
Document bug fix
Description:
IMPORTANT! IMPORTANT! IMPORTANT!
A case where metadata in a file could get corrupted in certain unusual
sitations was detected and fixed.
In certain circumstances, metadata could get cached in the raw data cache,
and if that particular piece of metadata was updated on disk while
incorrectly cached, the new metadata would get overwritten with the stale
metadata from the raw data cache when it was flushed out.
Additionally, I've patched up the raw data cache to be smarter about how
much it caches and how much I/O it triggers, leading to some speedups.
Solution:
Changed the raw data I/O routines which perform caching to require a
parameter with the size of the dataset being accessed and limited the
cache to no more than that many bytes.
Platforms tested:
FreeBSD 4.3 (hawkwind)
Update
Description:
Changed
#include <hdf_file.h>
construct to
#include "hdf_file.h"
so that the GNU compiler can more easily pick up the dependencies
which it places in the .depend and Dependencies files. Also
regenerated the Dependencies to go along with this.
Platforms tested:
Linux
Code cleanup.
Description:
Fixed _lots_ (I mean _tons_) of warnings spit out by the gcc with the
extra warnings. Including a few show-stoppers for compression on IRIX
machines.
Solution:
Changed lots of variables' types to more sensible and consistent types,
more range-checking, more variable typecasts, etc.
Platforms tested:
FreeBSD 4.2 (hawkwind), IRIX64-64 (modi4)
Feature symmetry
Description:
A while ago I needed to get the 'type' of data being accessed during writes
to the VFL driver, so I put in code to get the information down there.
Albert asked for the same information during reads, so I've added that in.
Tested:
FreeBSD 4.1.1 (hawkwind)
Maintainance & performance enhancements
Description:
Re-arranged header files to protect private symbols better.
Changed optimized regular hyperslab I/O to compute the offsets more
efficiently from previous method of using matrix operations.
Added sequential I/O operations at a more abstract level (at the same level
as H5F_arr_read/write), to support the optimized hyperslab I/O.
Platforms tested:
Solaris 2.6 (baldric) & FreeBSD 4.1.1 (hawkwind)
