GEMM-based level 3 BLAS: High-performance model implementations and performance evaluation benchmark

The level 3 Basic Linear Algebra Subprograms (BLAS) are designed to perform various matrix multiply and triangular system solving computations. Due to the complex hardware organization of advanced computer architectures the d evelopment of optimal level 3 BLAS code is costly and time consuming. Howev er, it is possible to develop a portable and high-performance level 3 BLAS library mainly relying on a highly optimized GEMM, the routine for the gene ral matrix multiply and add operation. With suitable partitioning all the o ther level 3 BLAS can be defined in terms of GEMM and a small amount of lev el 1 and level 2 computations. Our contribution is twofold. First, the mode l implementations in Fortran 77 of the GEMM-based level 3 BLAS are structur ed to reduce effectively data traffic in a memory hierarchy. Second, the GE MM-based level 3 BLAS performance evaluation benchmark. is a tool for evalu ating and comparing different implementations of the level 3 BLAS with the GEMM-based model implementations.