Portable high performance GEMM-based level 3 BLAS
Bo Kågström, Per Ling, Charles F. Van Loan · PPSC · 1993
The Level 3 Basic Linear Algebra subprograms (BLAS) are designed to perform various matrix multiply and triangular system solving computations. The development of optimal Level 3 BLAS code is costly and time consuming, because it requires assembly level programming/thinking. However, it is possible to develop a portable and high performance Level 3 BLAS only relying on an optimized GEMM, the BLAS subprogram for the general matrix multiply and add operation. With suitable partitioning, all the other Level 3 BLA subprograms can be defined in terms of GEMM and a negligible amount of Level 1 and 2 computations. Performance results of our portable GEMM-Based library for double precision real data are presented for various target architectures.