#define TS 32 #define WPT 16 #define RTS (int)(TS/WPT) // super super slow sgemm kernel by heehoon __kernel void sgemm(__global float *A, __global float *B, __global float *C, int M, int N, int K) { // Thread identifiers // Thread identifiers const int row = get_local_id(0); // Local row ID (max: TS/WPT == RTS) const int col = get_local_id(1); // Local col ID (max: TS) const int globalRow = TS*get_group_id(0) + row; // Row ID of C (0..M) const int globalCol = TS*get_group_id(1) + col; // Col ID of C (0..N) __local float Asub[TS][TS]; __local float Bsub[TS][TS]; float acc[WPT]; for (int w=0; w