Consider the following tiled matrix multiplication code:

#define TILE$\_$WIDTH

$\_\_$global$\_\_$ void MatrixMulKernel$($float$*$ d$\_$M$,$ float$*$ d$\_$N$,$ float$*$ d$\_$P$,$ int Width$)$

Question $2.(10$points$)$Consider the matrix multiplication with the input Matrix M and N $($M and N are both square matrixes, and the size of M and N is width $*$width$,$ here width is a power of $2,$ i$.$e$.,$ width $=2^$i$).$ How many times is each element in the input matrixes requested from global memory when:

$(1)$ There is no tiling.

$(2)$ Tiles of size T$*$T are used $($Suppose T is a power of $2,$ i$.$e$.,$ T$=2^$j$)$

$\{$

$01.\_\_$shared$\_\_$ float ds$\_$M$[$TILE$\_$WIDTH$][$TILE$\_$WIDTH$]$;

$02.\_\_$shared$\_\_$ float ds$\_$N$[$TILE$\_$WIDTH$][$TILE$\_$WIDTH$]$;

$03.$ int bx $=$ blockIdx.x; int by $=$ blockIdx.y;

$04.$ int tx $=$ threadIdx.x; int ty $=$ threadIdx.y;

$//$ Identify the row and column of the Pd element to work on

$05.$ int Row $=$ by $*$ TILE$\_$WIDTH $+$ ty;

$06.$ int Col $=$ bx $*$ TILE$\_$WIDTH $+$ tx;

$07.$ float Pvalue $=0$;

$//$ Loop over the Md and Nd tiles required to compute the Pd element

$08.$ for $($int m $=0$; m