v1.2.0

v1.2.0

• All lapack-related saf_utility_veclib functions now include an option to pre-allocate the run-time memory
• All saf_utility_veclib vector-vector operations, when using Apple Accelerate as the chosen performance library, now call the appropriate optimised routines. The Intel MKL vector-vector operations have been switched to the "mode overriding" variants, which allows some accuracy to be traded for faster computations
• Added some SIMD (SSE) optimised function alternatives, which may be enabled with "SAF_ENABLE_SIMD"
• A number of other minor and some significant optimisations throughout the framework
• Improved doxygen documentation for the examples
• Added resampleHRIRs() function to saf_hrir module, which employs the SPEEX resampler.

Breaking changes:

• Since the lapack-related saf_utility_veclib now allow a pre-allocated object to be passed to them, the first input argument should either be set to "NULL", (in which case they revert to how they were previously), or you can pass the pre-allocated object, (in which case no run--time malloc calls are made).
• The SAF_PRINT_VERSION_LICENSE_STRING macro has been removed; use the SAF_LICENSE_STRING instead.

v1.1.6

v1.1.6

• added a "spreader" example which demonstrates coherent and incoherent source spreading (default is binaural)
• added truncationEQ to saf_hoa module
• added Quaternion support to saf_utility_geometry.h/.c
• improvements to the image-source simulator in saf_reverb module, and the ambi_room_sim example
• improvements to the lattice decorrelator in saf_utility_decor.h/.c
• a number of bug fixes and improvements all round.

v1.1.5

v1.1.5

• Bug fix in the ACN/FuMa conversion.
• Some minor Doxygen improvements.
• Linked repository with zenodo.org to generate DOI for it.

v1.1.4

v1.1.4

• Doxygen-Documentation is now generated via a github-action and hosted here: https://leomccormack.github.io/Spatial_Audio_Framework/
• Added an improved SOFA reader, saf_sofa_open(), which fully conforms to the SOFA 1.0 FIR standard
• Added extras/matlab, which includes scripts to generate the default HRIR database, and the saf_sofa_open matlab equivalent.
• Added Intel IPP support to the saf_utility_fft.h/c wrapper. Enable with: "SAF_USE_INTEL_IPP".
• Further optimised the multi-channel and matrix convolvers in saf_utility_matrixConv.h/.c
• A number of bug fixes and improvements all round.

v1.1.3beta

v1.1.3beta

• Added an optional saf_tracker module (GPLv2) for tracking 3-D data points (e.g., direction-of-arrival estimates as unit vectors, Cartesian coordinates [x,y,z] etc.).
• A number of improvements to the image-source simulator in the saf_reverb module.
• Added CMake support for the safmex wrappers.

v1.1.2

v1.1.2

• Decoupled SAF from its external dependencies
• Added a extras/safmex folder which contains Matlab MEX wrappers; currently for the: afSTFT, QMF, IIR filterbank, VBAP gain tables, and getSHreal/complex SAF functions
• General bug fixes and improvements all round

Breaking changes:

• Due to the decoupling of SAF from its external dependencies, no longer will you be able to use SAF's dependencies within your own project by default; i.e. when including SAF via #include "saf.h", you will no longer also inherit the cblas/lapack functions etc. Instead, these external dependencies are included only internally within SAF by default. However, if you would like to retain access to these external dependencies within your own project, then you may now optionally include them by adding: #include "saf_externals.h" along with #include "saf.h".

v1.1.1

v1.1.1

• Reworked the afSTFT interface, and optimised its internals
• Added several more utility functions to saf_utilities, and added a resampler (based on libsamplerate) to resources
• Added more unit tests
• General bug fixes and improvements all round

Breaking changes:

• Due to the afSTFT interface rework, existing code that uses afSTFT will need to be modified. The main difference is that complexVector has been replaced with the native "float_complex" data type. It also no longer requires the input audio to be subdivided into chunks of "hopsize", as this is now done internally. Users must now also specify if they want the 3-D frequency-domain data permuted as: bands x channels x time, or time x channels x bands. The required changes should be quite self-explanatory, however, one may also refer to the examples (e.g. ambi_bin), which have been updated to use this new interface.

v1.1.0

v1.1.0

• Added comprehensive CMake support
• Documentation improvements
• Added more unit tests
• Added saf_utility_geometry.h/c (convexhull, spherical Delaunay + Voronoi functions etc.)
• General bug fixes and improvements all round

v1.0.0

First official release!