Release Notes#

This section includes significant changes, new features, performance improvements and various issues.

NVPL BLAS 0.4.0#

New Features#

  • Improved performance for small SGEMM on Arm Neoverse V2

  • Improved multihreaded performance for small-medium problem sizes on Arm Neoverse V2

  • Improved performance for GEMV with incx and incy not equal to 1

Known Issues#

  • N/A

Resolved Issues#

  • N/A

NVPL BLAS 0.3.0#

New Features#

  • N/A

Known Issues#

  • N/A

Resolved Issues#

  • N/A

NVPL BLAS 0.2.0#

New Features#

  • Improved performance on Arm Neoverse V2 for DGEMM when using small matrices

Known Issues#

  • N/A

Resolved Issues#

  • Valgrind crashes when run previous version of NVPL BLAS

NVPL BLAS 0.1.0#

The first early access release of NVPL BLAS library

New Features#

  • Standard C and Fortran BLAS APIs

  • Fortran 77 interfaces

  • Support lp64 and ilp64 interfaces

  • Support single- and multi-threaded (based on OpenMP) computation.

  • Extended API includes: - Batch GEMM API extensions: grouped batch and strided batch APIs - NVPL BLAS service functions: version and threading control

Known Issues#

  • N/A

Resolved Issues#

  • N/A