added version from workshop

cuda/vector_sum.cu

+#include <iostream>
+#include <vector>
+
+__global__ void sum_two_vectors(float* _device_a, 
+	   			float* _device_b,
+				unsigned int _size){
+				
+	unsigned int index = blockIdx.x*blockDim.x + threadIdx.x;
+	if(index < _size)
+		 _device_a[index] = _device_a[index] + _device_b[index];
+}
+
+int main(int argc, char** argv){
+
+    unsigned int vector_size = 1000*argc;
+    std::vector<float> host_a(vector_size,12.);
+    std::vector<float> host_b(vector_size,30.);	
+    std::cout << "created vectors of size " << vector_size << std::endl;
+
+    unsigned int vector_size_byte = vector_size*sizeof(float);
+    float* device_a = nullptr;
+    float* device_b = nullptr;	
+    cudaMalloc(&device_a,vector_size_byte);
+    cudaMalloc(&device_b,vector_size_byte);	
+    
+    cudaMemcpy(device_a,&host_a[0],vector_size_byte,cudaMemcpyHostToDevice);
+    cudaMemcpy(device_b,&host_b[0],vector_size_byte,cudaMemcpyHostToDevice);
+
+//work to be done on device_a and device_b
+    unsigned int threads = 32;//dim3
+    unsigned int blocks = (vector_size + threads -1 )/threads;
+
+    sum_two_vectors<<<blocks,threads>>>(device_a, device_b, vector_size);
+
+    cudaMemcpy(&host_a[0],device_a,vector_size_byte,cudaMemcpyDeviceToHost);
+    if(host_a[0]!=12.f)
+	std::cout << "GPU altered host_a\n";    
+
+    cudaFree(device_a);
+    cudaFree(device_b);
+    return 0;
+
+}
\ No newline at end of file