aidevnn / IntelGemm

GEneral Matrix Multiplication with Intel Compiler and his powerfull Autoparallelization and Autovectorization

IntelGemm

GEneral Matrix Multiplication with Intel Compiler and his powerfull Autoparallelization and Autovectorization

void MatMul0(int m, int n, int k, float* a, float* b, float* c) {
	for (int i = 0; i < m; ++i) {
		for (int j = 0; j < n; ++j) {
			for (int k0 = 0; k0 < k; ++k0) {
				c[i * n + j] += a[i * k + k0] * b[k0 * n + j];
			}
		}
	}
}

void MatMul1(int m, int n, int k, float* a, float* b, float* c) {
	for (int i = 0; i < m; ++i) {
		for (int k0 = 0; k0 < k; ++k0) {
			for (int j = 0; j < n; ++j) {
				c[i * n + j] += a[i * k + k0] * b[k0 * n + j];
			}
		}
	}
}

void MatMul2(int m, int n, int k, float* a, float* b, float* c) {
	cblas_sgemm(CblasRowMajor, CblasNoTrans, CblasNoTrans, m, n, k, 1.0f, a, k, b, n, 0.0f, c, n);
}

The Output

HelloWorld Intel MKL Gemm
Matrix[2 3] A
     0    -4     4
     7     2    -4

Matrix[3 4] B
    -3    -2    -3    -2
    -3     4    -9     3
   -10    -5     8    -3

Matrix[2 4] A x B meth0
   -28   -36    68   -24
    13    14   -71     4

Matrix[2 4] A x B meth1
   -28   -36    68   -24
    13    14   -71     4

Matrix[2 4] A x B meth2
   -28   -36    68   -24
    13    14   -71     4

Start Bench M=1920 N=2560 and K=1280
Bench method 0
TIME    97.63 ms
TIME    95.18 ms
TIME    91.10 ms
TIME    89.63 ms
TIME    88.26 ms
Bench method 1
TIME    88.22 ms
TIME    88.36 ms
TIME    89.07 ms
TIME    86.03 ms
TIME    88.58 ms
Bench method 2
TIME    89.54 ms
TIME    92.78 ms
TIME    91.48 ms
TIME    91.43 ms
TIME    92.41 ms
End.

Start Bench M=3840 N=5120 and K=2560
Bench method 0
TIME   675.68 ms
TIME   668.07 ms
TIME   674.81 ms
TIME   677.28 ms
TIME   673.65 ms
Bench method 1
TIME   669.31 ms
TIME   666.31 ms
TIME   673.02 ms
TIME   665.42 ms
TIME   696.23 ms
Bench method 2
TIME   704.99 ms
TIME   688.36 ms
TIME   675.51 ms
TIME   676.15 ms
TIME   677.55 ms
End.