Highlights of AOCL-BLAS 6.0.0
- New high-performance in-built allocator support
- Can be enabled using
--enable-aocl-alloc - Off by default
- Can be enabled using
- DGEMM performance tuning on Zen6 for large matrices
- C/ZGEMM optimizations on Zen4/5/6 for skinny/small matrices
- SGEMM optimizations on Zen5/6 for large/skinny/small matrices
- GEMV optimizations on Zen3/4/5/6
- GEMV kernels with multi-threading support added
- Multi-threading support added for GER
- New CBLAS imatcopy/omatcopy routines for in-place and out-of-place matrix copy and transpose
- Skip OpenMP parallel region for single-threaded L3 and SCALV operations
- Bug fixes and accuracy improvements