Installation

RandBLAS is a header-only C++20 library with two required dependencies. One of these dependencies (Random123) is header-only, while the other (BLAS++) needs to be compiled.

Having a compiled dependency makes setting up RandBLAS a little more complicated than setting up other header-only libraries. RandBLAS also has OpenMP and GoogleTest as optional dependencies. Access to OpenMP is essential for RandBLAS to achieve its best possible performance.

RandBLAS is most useful when called from programs that can access LAPACK, or an equivalent library for dense matrix computations. However, we don’t require that such a library is available.

CMake users

RandBLAS offers a CMake build system. This system manages a simple configuration step (populating RandBLAS/config.h), connecting to BLAS++ (assuming it was built with CMake), and building unit tests.

You can opt to install RandBLAS if you want other CMake projects to include it as a dependency. Formal installation just consists of copying the header files and CMake metadata into a directory of your choosing.

See INSTALL.md for detailed build and installation instructions. Check out our examples for CMake projects that use RandBLAS and LAPACK++ to implement high-level randomized algorithms.

Warning

Make sure to use the flag -Dblas_int=int64 in the CMake configuration line for BLAS++ If you don’t do that then you might get int32, which can lead to issues for large matrices.

Intel MKL (optional)

If BLAS++ was built against Intel MKL, then RandBLAS will automatically detect this at configure time and set RandBLAS_HAS_MKL in RandBLAS/config.h. Having MKL available provides two benefits:

  1. SpGEMM support. The RandBLAS::spgemm() function (sparse × sparse → dense) requires MKL. Without MKL, calling spgemm will produce a compile-time error. Note that spgemm only supports single and double precision (float and double), in contrast to other RandBLAS kernels that work with any scalar type.

  2. Accelerated SpMM. When MKL is present, RandBLAS::spmm() will use MKL’s optimized sparse BLAS routines for CSR and COO matrices (with a fallback to RandBLAS’ hand-rolled kernels for cases MKL cannot handle, such as CSC format or transposed dense operands).

No additional configuration is needed beyond ensuring BLAS++ was built with MKL.

If you want to use MKL for dense BLAS (through BLAS++) but disable RandBLAS’s MKL sparse features, pass -DRandBLAS_USE_MKL_SPARSE=OFF at configure time:

cmake -DRandBLAS_USE_MKL_SPARSE=OFF [other flags] ..

Everyone else

Strictly speaking, we only need three things to use RandBLAS in other projects.

  1. RandBLAS/config.h, filled according to the instructions in RandBLAS/config.h.in.

  2. The locations of Random123 header files.

  3. The locations of the header files and compiled binary for BLAS++ (which will referred to as “blaspp” when installed on your system).

If you have these things at hand, then compiling a RandBLAS-dependent program is just a matter of specifying standard compiler flags.

We recommend that you take a look at INSTALL.md. even if you aren’t using CMake, since it has additional advice about selecting an acceptable compiler or getting RandBLAS to see OpenMP.