High Performance Computing 1
How efficient is the BLAS?
• load/store float ops refs/ops
•level 1
•SAXPY 3N           2N   3/2
•level 2
•SGEMV MN+N+2M    2MN            1/2
•level 3
•SGEMM    2MN+MK+KN    2MNK            2/N