cyc_papi

    if ( mode == 2 ){//opltimal, ikj
        int lib = PAPI_library_init(PAPI_VER_CURRENT);
        if ( lib == 0 ) cout << "lib == 0\n";
        const PAPI_hw_info_t *info = PAPI_get_hardware_info();
        if (!info)
            cout << "PAPI_get_hardware_info() returns NULL\n";
        int blocksize = sqrt((info->mem_hierarchy.level[0].cache[0].size)/3);
//      cout << blocksize << endl;
        for ( i = 0; i < rows; i += blocksize ){
            for ( k = 0; k < columns; k += blocksize ){
                for ( j = 0; j < M2.columns; j += blocksize ){
                //block mult
                    for ( int i1 = i; i1 < i + blocksize && i1 < rows; i1++ ){
                        for ( int k1 = k; k1 < k + blocksize && k1 < M2.columns; k1++ ){
                            for ( int j1 = j; j1 < j + blocksize && j1 < columns; j1++ ){
                                Tmp.vals[i1][j1] += vals[i1][k1]*M2.vals[k1][j1];
                            }
                        }
                    }
                }
            }
        }
    }