HDF5 2.0 Planning

Jump to bottom

Dana Robinson edited this page Oct 30, 2024 · 10 revisions

These are the topics/issues we're going to focus on in the next version of HDF5. This list is ambitious, so not everything here is realistically going to make it in.

NOTE: This list is being adapted from an earlier HDF5 1.16.0 to-do list. It should be finalized by early November (for a March 2025 release).

Tentative 2.0.0 changes

Update library defaults to provide better performance with cloud-optimized HDF5 and modern I/O hardware
Complex number support
Build system changes
- Better alignment between Autotools and CMake
- Might drop/rename some build features
Use HDoff_t instead of off_t in public API

Bugfixes and general quality

Fix all oss-fuzz and other segfault-related bugs (there are several in Jira) and harden surrounding code
-fsanitize=memory/address/undefined clean and added to CI
Fix file image bugs and review code (#1915)

Cloud-Optimized HDF5

Review and possibly refactor ros3 VFD code
Performance testing

Support for new datatypes

Complex numbers
ML types (FP8, FP4)
Fix outstanding long double issues (mainly POWER)

Configure issues

Autotools pkg-config support
CMake support for MinGW + MSYS2

Performance

Additional subfiling performance profiling and fixing

Windows as a first-class citizen

Win32 VFD
Unicode/code page file name tests
H5Dwrite performance issue
Core VFD slowdown

CI/CD/Testing

Determine a final version naming/numbering scheme that works well with package management systems
Clean up CDash, public CDash all green
Schedule a pre-release and code freeze
Fix MacOS signed binary issues
Develop a manual test process
Move VOL tool testing to main repo

Documentation

All “developer-level” public headers have full Doxygen (VOL, VFD, etc.)
Develop a list of user guide improvements that need to be made

Misc

Close PR #1387 (minor datatype optimization)
Go over plugin, etc. paths and make sure that everything is handled well and uniformly

Tools

H5L copy hack in h5repack

Fortran wrappers

Make sure cross-compiling with Fortran works well