fix the bug in local_max about casting size_t to int

Laila Saadatifard
1 parent 7bca968e
Showing 1 changed file with 6 additions and 20 deletions Show diff stats
stim/iVote/ivote2/local_max.cuh
@@ -10,30 +10,22 @@ namespace stim{
 		// this kernel calculates the local maximum for finding the cell centers
 		template<typename T>
-		__global__ void cuda_local_max(T* gpuCenters, T* gpuVote, int conn, size_t x, size_t y){
+		__global__ void cuda_local_max(T* gpuCenters, T* gpuVote, int conn, int x, int y){
 			// calculate the 2D coordinates for this current thread.
-			size_t xi = blockIdx.x * blockDim.x + threadIdx.x;
-			size_t yi = blockIdx.y * blockDim.y + threadIdx.y;
+			int xi = blockIdx.x * blockDim.x + threadIdx.x;
+			int yi = blockIdx.y * blockDim.y + threadIdx.y;
 			if(xi >= x || yi >= y)
 				return;
 			// convert 2D coordinates to 1D
-			size_t i = yi * x + xi;
+			int i = yi * x + xi;
 			gpuCenters[i] = 0;		//initialize the value at this location to zero
 			T val = gpuVote[i];
-			//compare to the threshold
-			//if(val < final_t) return;
-			
-			//define an array to store indices with same vote value
-			/*int * IdxEq;
-			IdxEq = new int  [2*conn];
-			int n = 0;*/
-			
 			for(int xl = xi - conn; xl < xi + conn; xl++){
 				for(int yl = yi - conn; yl < yi + conn; yl++){
 					if(xl >= 0 && xl < x && yl >= 0 && yl < y){
@@ -42,8 +34,7 @@ namespace stim{
 							return;
 							}
 						if (gpuVote[il] == val){
-							/*IdxEq[n] = il;
-							n = n+1;*/
+							
 							 if( il > i){
 								 return;
 							}
@@ -51,12 +42,7 @@ namespace stim{
 					}							
 				}
 			}
-			/*if (n!=0){
-				if(IdxEq[n/2] !=i){
-					return;
-				}
-			}	*/	
-			//gpuCenters[i] = 1;
+			
 			gpuCenters[i] = gpuVote[i];
 		}