代码

#include <iostream>
#pragma GCC target("avx2,fma")

using namespace std;
template<size_t L,size_t M,size_t N>
struct MatMulKernel
{
    enum
        {
        row1=L,
        col1=M,
        row2=M,
        col2=N
    };
    template<typename Ty>
    static void matmul(const Ty *A,const Ty *B,Ty *C)
    {
        for(size_t i=0;i<row1;i++)
        {
            for(size_t j=0;j<col1;j++)
            {
                Ty A_ij=A[i*col1+j];
                for(size_t k=0;k<col2;k++)
                {
                    C[i*col2+k]+=B[j*col2+k]*A_ij;
                }
            }
        }
    }
};
void matrix_multiply(int n, const double *A, const double *B, double *C) {
	MatMulKernel<1024,1024,1024>::matmul(A,B,C);
}

评测结果