Added run script, fix for cpu nw benchmark

Fixed selection and harvesting of the OCL platform and changed the
hardcoded variable to select the CPU device

opencl/nw/nw.c

@@ -64,10 +64,17 @@ static int initialize(int use_gpu)
 	cl_int result;
 	size_t size;
 
+	// modification to handle the case in which we have more than one OpenCL platform available on the system.
+	cl_uint platformCount;
+	
+
 	// create OpenCL context
-	cl_platform_id platform_id;
-	if (clGetPlatformIDs(1, &platform_id, NULL) != CL_SUCCESS) { printf("ERROR: clGetPlatformIDs(1,*,0) failed\n"); return -1; }
-	cl_context_properties ctxprop[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)platform_id, 0};
+	clGetPlatformIDs(0, NULL, &platformCount);
+	
+	cl_platform_id *platforms_ids;
+	platforms_ids = (cl_platform_id*) malloc(sizeof(cl_platform_id) * platformCount);
+	if (clGetPlatformIDs(platformCount, platforms_ids, NULL) != CL_SUCCESS) { printf("ERROR: clGetPlatformIDs(1,*,0) failed\n"); return -1; }
+	cl_context_properties ctxprop[] = { CL_CONTEXT_PLATFORM, (cl_context_properties)platforms_ids[1], 0};
 	device_type = use_gpu ? CL_DEVICE_TYPE_GPU : CL_DEVICE_TYPE_CPU;
 	context = clCreateContextFromType( ctxprop, device_type, NULL, NULL, NULL );
 	if( !context ) { printf("ERROR: clCreateContextFromType(%s) failed\n", use_gpu ? "GPU" : "CPU"); return -1; }
@@ -223,7 +230,7 @@ int main(int argc, char **argv){
 	size_t local_work[3] = { (workgroupsize>0)?workgroupsize:1, 1, 1 };
 	size_t global_work[3] = { nworkitems, 1, 1 }; //nworkitems = no. of GPU threads
 	
-	int use_gpu = 1;
+	int use_gpu = 0;
 	// OpenCL initialization
 	if(initialize(use_gpu)) return -1;
 

opencl/nw/run-cpu

@@ -0,0 +1 @@
+./nw 6400 10 ./nw.cl