Releases: eth-cscs/SpFFT
Releases · eth-cscs/SpFFT
SpFFT 1.1.0
SpFFT 1.1.0 Release Notes
Changes
- Switch to C++17
- CUDA version requirement increased to 11.0
- CMake version requirement increased to 3.18
- Updated dependencies required to build tests
- Added CMake option to disable downloading of test dependencies
Bug Fixes
- Fixed compilation with ROCm 6.0
- Fixed issue where building in release mode with ROCm did not disable debug checks, leading to decreased performance
SpFFT 1.0.6
SpFFT 1.0.6 Release Notes
Bug Fixes
- Added workaround for compiler bug in ROCm 4.5 and later, causing missing kernel symbols
SpFFT 1.0.5
SpFFT 1.0.5 Release Notes
Features
- Support for Arm Performance Libraries
Bug Fixes
- Fixed a potential performance issue on AMD GPUs with latest versions of ROCm / HIP due to changes required for device pointer detection
- Fixed an issue with R2C transform without local data
SpFFT 1.0.4
SpFFT 1.0.4 Release Notes
Bug Fixes
- Linking to single precision FFTW library is now only done when actually required
- Fixed issues with installed CMake config files:
- Custom find modules are now found correctly for static library
- Workaround for bug in find_dependency macro with CMake < 3.15.0, where components of MPI and OpenMP may be missing.
SpFFT 1.0.3
SpFFT 1.0.3 Release Notes
Bug Fixes
- Fixed missing header include, causing a compilation error with GCC 11 in some cases
SpFFT 1.0.2
SpFFT 1.0.2 Release Notes
- ROCm 4.1.0 support
SpFFT 1.0.1
SpFFT 1.0.1 Release Notes
Bug Fixes
- Fixed incorrect memory transfer when executing on GPU, which caused an error when running with Nvidia Hyper-Q enabled
SpFFT 1.0.0
SpFFT 1.0.0 Release Notes
This is a new major release, introducing new functionalities.
It is fully backward compatible to previous release, but includes extensions to the API.
Major Changes
- Support for externally allocated memory for space domain data including in-place and out-of-place transforms
- Optional asynchronous computation when using GPUs
- Simplified / direct transform handle creation if no resource reuse through grid handles is required
Minor Changes
- Added CMake option to allow explicit selection of library providing a FFTW3 implementation
- Updated ROCm build to use installed CMake config files
- SpFFT CMake config file generated upon installation is now fully relocatable for static builds as well
SpFFT 0.9.13
SpFFT 0.9.13 Release Notes
Bug fixes
- Added workaround for bug in rocFFT for sizes (1, 1, z)
- Added workaround for bug in ROCm 3.5 (and later), where device pointers with offset from start of allocated memory were not detected correctly
SpFFT 0.9.12
SpFFT 0.9.12 Release Notes
Bug fixes
- Added workaround for bug in cuFFT introduced with CUDA 10.2 for some edge cases of size (1, y, z), where y is a prime number
- Fixed detection of device pointers with CUDA 11