What type of incorrect execution

behavior can happen if one forgot to

use syncthreads () in

the kernel of Figure 4.16?

_global_ void Matrixul Kernel (float* dum, float* dux, float. a_P. int Width) 1. 2. 3. 5. 6. 7. 8. _shared_float MdS TILE_WIDTH] TILE_WIDTH) _shared_float Nas [TILE_WIDTH TILE_WIDTH) int bx = blockix.x; int by - blockIdx.y: int tx = threadIdx.x; int ty = threadIdx.y: // Identify the row and column of the dP element to work on int Row - by . TILR_WIDTH - ty: int Col - bx. TILE_WIDTH. Ex: float Pvalue - 0 // Loop over the duM and eN tiles required to compute d P element for (int ph = 0; ph