OpenBLAS 0.3.4 Released With Intel AVX-512 Optimizations, Other Performance Tuning
For those utilizing OpenBLAS as the linear algebra library for your application(s), OpenBLAS 0.3.4 was released on Sunday with the latest features and CPU optimizations for multiple architectures.
OpenBLAS 0.3.4 now provides enough buffer space for at least 50 threads by default, AMD Ryzen 2 auto-detection has been fixed, there are Intel AVX-512 optimizations yielding up to ~2x performance improvements, improved IBM POWER CPU detection, and various ARM improvements.
More information on the many changes to find with the OpenBLAS 0.3.4 linear algebra library can be found via OpenBLAS on GitHub.
OpenBLAS 0.3.4 now provides enough buffer space for at least 50 threads by default, AMD Ryzen 2 auto-detection has been fixed, there are Intel AVX-512 optimizations yielding up to ~2x performance improvements, improved IBM POWER CPU detection, and various ARM improvements.
More information on the many changes to find with the OpenBLAS 0.3.4 linear algebra library can be found via OpenBLAS on GitHub.
2 Comments