13.1.2. Building from Source on Linux - 13.1.2. Building from Source on Linux - 5.3 English - 57404

Complete the following steps to build AOCL-LibMem for Linux:

1. Download and install the AOCL master installer (aocl-linux-<compiler>-<version>.tar.gz) from:

   https://www.amd.com/en/developer/aocl.html

2. Locate the aocl-libmem folder in the root directory.

3. Configure for one of the following builds as required:

   • GCC build

     Default Option - Compiling for Native CPU

     $ cmake -D CMAKE_C_COMPILER=gcc -S <source_dir> -B <build_dir>
     

     Compiling for Specific Zen Architecture (zen1, zen2, zen3, zen4, zen5)

     $ cmake -D CMAKE_C_COMPILER=gcc -D ALMEM_ARCH=zen4 -S <source_dir> -B <build_dir>
     

     Compiling with Dynamic CPU Dispatcher

     $ cmake -D CMAKE_C_COMPILER=gcc -D ALMEM_DYN_DISPATCH=Y -S <source_dir> -B <build_dir>
     

     Compiling AVX2 binary

     $ cmake -D CMAKE_C_COMPILER=gcc -D ALMEM_ISA=avx2 -S <source_dir> -B <build_dir>
     

     Compiling AVX512 binary

     $ cmake -D CMAKE_C_COMPILER=gcc -D ALMEM_ISA=avx512 -S <source_dir> -B <build_dir>
     

     Compiling with Tunable Parameters

     $ cmake -D CMAKE_C_COMPILER=gcc -D ALMEM_TUNABLES=Y -S <source_dir> -B <build_dir>
     

   • AOCC (Clang) build

     Default Option - Compiling for Native CPU

     $ cmake -D CMAKE_C_COMPILER=clang -S <source_dir> -B <build_dir>
     

     Compiling for Specific Zen Architecture (zen1, zen2, zen3, zen4, zen5)

     $ cmake -D CMAKE_C_COMPILER=clang -D ALMEM_ARCH=zen4 -S <source_dir> -B <build_dir>
     

     Compiling with Dynamic CPU Dispatcher

     $ cmake -D CMAKE_C_COMPILER=clang -D ALMEM_DYN_DISPATCH=Y -S <source_dir> -B <build_dir>
     

     Compiling AVX2 binary

     $ cmake -D CMAKE_C_COMPILER=clang -D ALMEM_ISA=avx2 -S <source_dir> -B <build_dir>
     

     Compiling AVX512 binary

     $ cmake -D CMAKE_C_COMPILER=clang -D ALMEM_ISA=avx512 -S <source_dir> -B <build_dir>
     

     Compiling with Tunable Parameters

     $ cmake -D CMAKE_C_COMPILER=clang -D ALMEM_TUNABLES=Y -S <source_dir> -B <build_dir>
     

   Compiling Strict-Bounds Variant

   $ cmake -D CMAKE_C_COMPILER=gcc -D ALMEM_ISA=avx512 -D ALMEM_STRICT_BOUNDS=Y -S <source_dir> -B <build_dir>
   

   Note

   ‘ALMEM_ARCH’ has highest precedence, followed by ‘ALMEM_ISA’, then ‘ALMEM_DYN_DISPATCH’ and ‘ALMEM_TUNABLES’ options.

   Note

   ‘ALMEM_STRICT_BOUNDS’ can be combined with any of the above build modes. By default, the library uses page-boundary-safe unmasked vector loads for small-size code paths to improve throughput. The strict-bounds variant replaces these with masked loads that never read beyond the caller-specified buffer size, even within the same mapped page. The tradeoff is a slight reduction in performance, as masked loads have higher latency than unmasked loads.

4. Build:

   $ cmake --build <build_dir>
   

5. Install:

   $ cmake --install <build_dir>
   
   ## For custom install path, run configure with "CMAKE_INSTALL_PREFIX"