Merge pull request #1986 from peastman/error

Fixed compilation error in kernel

platforms/cuda/src/kernels/rmsd.cu

@@ -6,6 +6,7 @@
  */
 __device__ real reduceValue(real value, volatile real* temp) {
     const int thread = threadIdx.x;
+    __syncthreads();
     temp[thread] = value;
     __syncthreads();
     for (uint step = 1; step < 32; step *= 2) {

platforms/opencl/src/kernels/rmsd.cl

@@ -6,12 +6,13 @@
  */
 real reduceValue(real value, __local volatile real* temp) {
     const int thread = get_local_id(0);
+    barrier(CLK_LOCAL_MEM_FENCE);
     temp[thread] = value;
     barrier(CLK_LOCAL_MEM_FENCE);
     for (uint step = 1; step < 32; step *= 2) {
         if (thread+step < get_local_size(0) && thread%(2*step) == 0)
             temp[thread] = temp[thread] + temp[thread+step];
-        SYNC_WARPS
+        SYNC_WARPS;
     }
     for (uint step = 32; step < get_local_size(0); step *= 2) {
         if (thread+step < get_local_size(0) && thread%(2*step) == 0)