AMD Optimizing CPU Libraries (AOCL)

NEW! AOCL 5.1 is now available, May 7, 2025

AOCL is a set of numerical libraries optimized for AMD processors based on the AMD “Zen” core architecture and generations. Supported processor families are AMD EPYC™, AMD Ryzen™, and AMD Ryzen™ Threadripper™ processors. The tuned implementations of industry-standard math libraries enable rapid development of scientific and high-performance computing applications.

Libraries

• AOCL-Data Analytics
• AOCL-BLAS
• AOCL-LAPACK
• AOCL-Compression
• AOCL-Cryptography
• AOCL-FFTW (Fastest Fourier Transform in the West)
• AOCL-LibM (AMD Math Library)
• AOCL-LibMem
• AOCL-RNG (AMD Random Number Generator Library)
• AOCL-SecureRNG (Secure RNG Library)
• AOCL-ScaLAPACK
• AOCL-Sparse
• AOCL-Utils

What’s new in AOCL 5.1 (May 7, 2025)

• AOCL Build-It-Yourself (Beta)
  • AOCL now offers the capability to compile individual libraries and consolidate them into a unified binary. With the Build-It-Yourself feature, you can choose one or more AOCL libraries and merge them into a single library by configuring the appropriate CMake options.
  • This unified binary is assigned a default name: libaocl.so/ libaocl.a for Linux and aocl.dll/ aocl.lib for Windows
  • This approach simplifies integration by eliminating dependencies on library linking order and preventing API duplication, ensuring smooth and efficient incorporation of multiple AOCL libraries.
• AOCL-Data Analytics 
  • New APIs for DBSCAN clustering, support vector machines (SVM) and pairwise distances.
  • Dynamic dispatch: multiple Zen generation code paths compiled in a single binary, with appropriate code path selected at runtime.
  • Performance improvements to decision trees, random forest and k-nearest neighbors algorithms.
• AOCL-BLAS
  • Performance Optimizations
    • DGEMM, DTRSM, DGEMV, ZGEMM, DTRSV, DCOPYV on Zen4/5
    • DSCALV, DDOTV on Zen3
  • Benchmark support for ASUMV
  • Aocl-gemm Add-on Module updates
    • AOCL_ENABLE_INSTRUCTIONS support
    • batch_gemm support for all data types
    • New Output Datatype for Integer APIs
    • BF16 Support on AVX2 Platforms
    • WOQ with/without Group Quantization
    • Threading Framework Optimizations
    • Reference Kernels for all Reorder APIs
    • Performance Optimizations for all APIs
    • Additional APIs and Post-Ops support in addition to the improved performance for the existing APIs in aocl_gemm add-on.
• AOCL-Compression
  • Introduced multithreaded APIs for GZIP and Raw Deflate formats.
  • Improved single threaded compression performance for Bzip2, Snappy and decompression performance for Bzip2, Snappy, Gzip, ZSTD.
  • AOCL_DECOMPRESS_FAST support extended to Snappy.
  • Integrated third party utilities: minigzip and zstd utility for benchmarking.
• AOCL-Cryptography
  • Performance improvements in SHA3.
  • AES-GCM optimizations for real world applications.
  • OpenSSL provider support for SHA3 and SHAKE algorithms.
  • Runtime forcing of CPUID / micro arch code path support using aocl-utils
  • CMake preset support.
  • Bench and test framework improvements.
  • Coverity high and majority of medium severity defects fixed.
• AOCL-FFTW
  • Same as the 5.0 release.
• AOCL-LAPACK
  • Performance improvements in the following routines:
  • • LU Factorizations (DGETRF & DGBTRF), Least Square Solver (ZGELS)
    • SVD (DGESDD) and Eigen Decomposition (DSYEVD)
  • Upgrade to Netlib LAPACK 3.12.0 specifications
  • Updates to CMAKE presets, pkg-config for ease of build and integration
  • Test cases added for C++ template interfaces
• AOCL-LibM
  • Added new vector variants of sincos.
  • Improved performance of vector variants of linearfrac.
  • Added CMAKE support (experimental) for building AOCL-LibM binaries on Linux and Windows.
  • Added static dispatch functionality while building with CMAKE.
  • Updated dynamic dispatch to route to Zen3 versions of APIs on Zen5 with AVX512 ISA disabled.
• AOCL-LibMem
  • Optimized memory functions for ZEN5 microarchitecture.
  • Optimized string functions for AVX512 and AVX2
  • Enabled support for Dynamic Dispatcher
  • Support for new string functions: strchr, strncat
  • Bench and Test Framework Improvements
• AOCL-RNG
  • AVX512 support for double precision MRG32K3A Generator.
  • Updated the AOCL-SecureRNG library to adjust the retrial count.
• AOCL-ScaLAPACK
  • Improved PDSYEVD performance for flat grid configurations
  • pkgconfig support for better integration with AOCL dependencies
• AOCL-Sparse
  • Multithreaded support for CSRMM, SpMV Transposed
  • Performance improvements:
    • Level 2: SpMV variants (Hermitian, Symmetric, Transposed)
    • Level 3: CSRMM
  • Enable dynamic dispatch
  • Support for complex data types in iterative methods, ILU0
  • Support for CSC storage format in TRSV
  • Framework improvements
  • Build system enhancements:
    • Enable CMake presets
    • Allow separate build of tests
• AOCL-Utils
  • Implemented Hardware and System Rng.
  • Implemented extendable and flexible thread safe logger.
  • Support for isZenFamily.
• AOCL-Enabled MUMPS Library
  • Building for Windows 10 systems: CMake-based build system on AOCL-enabled MUMPS sparse solver library; refer to GitHub MUMPS-build (https://github.com/amd/mumps-build) for more information.
  • Mumps 5.5.1 support with 5.5.0 and 5.4.1 backward compatibility.

Download with End User License Agreement

Resources and Technical Support

Documentation

• AOCL User Guide
• AOCL Tuning Guide
• AOCL Release Notes
• AOCL API Guide
• AOCL Build-It-Yourself Source code: GitHub
• Prior versions: AOCL Archive.

Support

For support options, refer to Technical Support.

AMD Community

For moderated forums, refer to the AMD community.