代码

#define idx(i, j) ((i) * n + (j))

#define expect(x, v) __builtin_expect(x, v)
#define likely(x) expect(!!(x), 1)
#define unlikely(x) expect(!!(x), 0)

#define unreachable() __builtin_unreachable()
#define assume(x) ({ if (!(x)) unreachable(); })

#define prefetch __builtin_prefetch

void matrix_multiply(int n, const double *A, const double *B, double *C) {
	for (int i = 0; i < n; i++)
		for (int j = 0; j < n; j++)
			C[idx(i, j)] = 0;
	for (int i = 0; i < n; i++) {
prefetch(C, idx(i, 0), 1);
		for (int k = 0; k < n; k++) {
prefetch(A + idx(i, k), 0);
prefetch(B + idx(k, 0), 0)
		for (int j = 0; j < n; j++) {
				C[idx(i, j)] += A[idx(i, k)] * B[idx(k, j)];
			}
		}
	}
}

评测结果