CODEWITHSUNDEEP
cuda2dsummultiplication
asma
asma

Reputation: 113

Sum of Multiplication for 2D array in CUDA

I'm new to CUDA , and I want to implement a sum of multiplication as this equation

enter image description here

enter image description here

I wrote this code in CUDA , but it doesn't give the correct answer

mulFV1[idx] = f[idx][idy]*compV2[idy];
mulFV2[idy] = f[idx][idy]*compV1[idx];

and then , I send the the arrays mulFV1 and mulFV2 to a reduction device function..

The question is how can I debug it?

Note :To be in the picture mulFV1 is concern in the rows and mulFV2 concern in the columns

Upvotes: 0

Views: 1084

Answers (2)

Yappie
Yappie

Reputation: 399

I think, that you kernel may be look like this following

__global__ void kernel_code(const int* f,const int* v1,const int* v2, int* outv1, int* outv2)
{
    int idx = blockIdx.x * blockDim.x + threadIdx.x;
    int idy = blockIdx.y * blockDim.y + threadIdx.y;
    if (idx<MAX_X && idy <MAX_Y)
    {
        if(idx==0)
        {
            outv2[idy]=0;
        }
        if(idy==0)
        {
            outv1[idx]=0;
        }
        __syncthreads();

        atomicAdd(&(outv1[idx]),f[idy*MAX_Y+ idx]*v2[idy]);
        atomicAdd(&(outv2[idy]),f[idy*MAX_Y+idx]*v1[idx]);
   }
}

Upvotes: 1

Howard
Howard

Reputation: 39197

Your variable names indicate that the first line is the multiplication using vector v1 and the second with v2. But instead you are doing it cross-over. Maybe you want to have

mulFV1[idx] = f[idx][idy]*compV1[idy];
mulFV2[idy] = f[idx][idy]*compV2[idx];

with indices 1 and 2 exchanged?

Upvotes: 0

Related Questions