Increase L2 and L3 cache size for Power10.
By increasing the L2 and L3 sizes for Power10, situations which rely on breaking matrices into sub-matrices (like triangular matrix solve) for packing and GEMM, the performance increases by 1.33X.
By increasing the L2 and L3 sizes for Power10, situations which rely on breaking matrices into sub-matrices (like triangular matrix solve) for packing and GEMM, the performance increases by 1.33X.