Be notified of new releases
Create your free GitHub account today to subscribe to this repository for new releases and build software alongside 40 million developers.Sign up
- Further fixes to
pthreadlinking problems for Linux compilation under GCC;
- Fixed issue with odd state counts and CPU SSE;
- Fixed issue with non nucleotide state counts on AMD GPUs;
- Adjustments to CPU auto-threading heuristics.
pthreadlinking issue on Linux which prevented some programs from loading CPU plugin.
- Fixed issue with multi-partition mode and unrooted trees on CPU implementation (multi-partition on CPU typically only used for debugging)
- Automatic resource selection via
- Control of CPU threading with
Performance and bug Fixes:
- CPU implementation now defaults to non-threaded implementation to replicate behaviour of previous versions.
- Improved CPU auto-threading to use fewer threads for better performance in some scenarios.
- Improved performance for CUDA implementation when rescaling (improved synchonization of streams).
- Fixed macOS installer to place BEAGLE JNI library in correct location
- Removed OpenCL implementation from macOS build and installer, as OpenCL is being deprecated by Apple
- Improved performance for partitioned data sets.
- Improved performance for large trees with short alignments.
- Improved performance for default CPU implementation, via use of threading.
- Several minor fixes.