4 ms·
Most decent BLAS implementations use some variation of the kernels in K. Goto et al. "Anatomy of High-Performance Matrix Multiplication" written in assembly.
by zzzoom 3y ago
Most decent BLAS implementations use some variation of the kernels in K. Goto et al. "Anatomy of High-Performance Matrix Multiplication" written in assembly.