Module SGBL3B in BLASGEMM-Based Level 3 BLAS (single precision model implementation). Basic matrix-matix operations such as multiplication, rank k updates and solving triangular systems. Designed to be highly efficient (i.e. to reduce data traffic) on machines with a memory hierarchy, e.g. a single processor on machines with local or global caches or micro processors with on-chip caches. Can be parallelized. (Developed by Bo Kagstrom, Per Ling and Charles Van Loan).