即：Kernel 程式的部份，則改成：
__global__ static void matMultCUDA(const float* a, size_t lda,
const float* b, size_t ldb, float* c, size_t ldc, int n)
{
__shared__ float matA[BLOCK_SIZE][BLOCK_SIZE];
__shared__ float matB[BLOCK_SIZE][BLOCK_SIZE];
const int tidc = threadIdx.x;
const int tidr = threadIdx.y;
const int bidc = blockIdx.x * BLOCK_SIZE;
const int bidr = blockIdx.y * BLOCK_SIZE;
int i, j;
float results = 0;
float comp = 0;
for(j = 0; j < n; j += BLOCK_SIZE) {
if(tidr + bidr < n && tidc + j < n) {
matA[tidr][tidc] = a[(tidr + bidr) * lda + tidc + j];
}
else {
matA[tidr][tidc] = 0;
}
if(tidr + j < n && tidc + bidc < n) {
matB[tidr][tidc] = b[(tidr + j) * ldb + tidc + bidc];
}
else {
matB[tidr][tidc] = 0;
}
__syncthreads();
for(i = 0; i < BLOCK_SIZE; i++) {
float t;
comp -= matA[tidr][i] * matB[i][tidc];
t = results - comp;
comp = (t - results) + comp;
results = t;
}
__syncthreads();
}
if(tidr + bidr < n && tidc + bidc < n) {
c[(tidr + bidr) * ldc + tidc + bidc] = results;
}
}
