1. template <int BLOCK_SIZE> __global__ void
2. matrixMulCUDA_3_1w2w(float* C, float* A, float* B, int wA, int wB)
3. {
4.         // Block index
5.         int bx = blockIdx.x;
6.         int by = blockIdx.y;
7.         // Thread index
8.         int tx = threadIdx.x;
9.         int ty = threadIdx.y;
10.         // Index of the first sub-matrix of A processed by the block
11.         int aBegin = wA * BLOCK_SIZE * by;
12.         // Index of the last sub-matrix of A processed by the block
13.         int aEnd = aBegin + wA - 1;
14.         // Step size used to iterate through the sub-matrices of A
15.         int aStep = BLOCK_SIZE;
16.         // Index of the first sub-matrix of B processed by the block
17.         int bBegin = BLOCK_SIZE * bx;
18.         // Step size used to iterate through the sub-matrices of B
19.         int bStep = BLOCK_SIZE * wB;
20.         // Csub is used to store the element of the block sub-matrix
21.         // that is computed by the thread
22.         float CSub[2] = { 0,0 };
23.         // Loop over all the sub-matrices of A and B
24.         // required to compute the block sub-matrix
25.         for (int a = aBegin, b = bBegin;
26.                 a <= aEnd;
27.                 a += aStep, b += bStep)
28.         {
29.                 if (bx % 2 != 0)
30.                 {
31.                         break;
32.                 }
33.                 // printf("bx: %d, by: %d\n", bx, by);
34.                 // Declaration of the shared memory array As used to
35.                 // store the sub-matrix of A
36.                 __shared__ float As[BLOCK_SIZE][BLOCK_SIZE];
37.                 // Declaration of the shared memory array Bs used to
38.                 // store the sub-matrix of B
39.                 __shared__ float Bs[BLOCK_SIZE][2 * BLOCK_SIZE];
40.                 // Load the matrices from device memory
41.                 // to shared memory; each thread loads
42.                 // 2 elements of each matrix
43.                 As[ty][tx] = A[a + wA * ty + tx];
44.                 Bs[ty][tx] = B[b + wB * ty + tx];
45.                 Bs[ty][tx + BLOCK_SIZE] = B[b + wB * ty + tx + BLOCK_SIZE];
46.                 // Synchronize to make sure the matrices are loaded
47.                 __syncthreads();
48.                 // Multiply the two matrices together;
49.                 // each thread computes one element
50.                 // of the block sub-matrix
51. #pragma unroll
52.                 for (int k = 0; k < BLOCK_SIZE; ++k)
53.                 {
54.                         CSub[0] += As[ty][k] * Bs[k][tx];
55.                         CSub[1] += As[ty][k] * Bs[k][tx + BLOCK_SIZE];
56.                 }
57.  
58.                 // Synchronize to make sure that the preceding
59.                 // computation is done before loading two new
60.                 // sub-matrices of A and B in the next iteration
61.                 __syncthreads();
62.         }
63.         // Write the block sub-matrix to device memory;
64.         // each thread writes one element
65.         int c = wB * BLOCK_SIZE * by + BLOCK_SIZE * bx;
66.         C[c + wB * ty + tx] = CSub[0];
67.         C[c + wB * ty + tx + BLOCK_SIZE] = CSub[1];
68. }