Fixed issue that caused inefficient sorting when a block contained only one atom (#5215)

* Fixed issue that caused inefficient sorting when a block contained only one atom

* Add the fix to OpenCL and HIP

platforms/cuda/src/kernels/findInteractingBlocks.cu

@@ -109,6 +109,7 @@ extern "C" __global__ void computeSortKeys(const real4* __restrict__ blockBoundi
         sizeRange = blockSizeRange[0];
         for (int i = 1; i < numSizes; i++) {
             real2 size = blockSizeRange[i];
+            if (size.x > 0)
                 sizeRange.x = min(sizeRange.x, size.x);
             sizeRange.y = max(sizeRange.y, size.y);
         }

platforms/hip/src/kernels/findInteractingBlocks.hip

@@ -115,6 +115,7 @@ extern "C" __global__ void findBlockBounds(int numAtoms, real4 periodicBoxSize,
 
         // Record the range of sizes.
         real totalSize = blockSize.x+blockSize.y+blockSize.z;
+        if (totalSize > 0)
             atomicMin(&blockSizeRange->x, totalSize);
         atomicMax(&blockSizeRange->y, totalSize);
     }

platforms/opencl/src/kernels/findInteractingBlocks.cl

@@ -75,6 +75,7 @@ __kernel void computeSortKeys(__global const real4* restrict blockBoundingBox, _
         sizeRange = blockSizeRange[0];
         for (int i = 1; i < numSizes; i++) {
             real2 size = blockSizeRange[i];
+            if (size.x > 0)
                 sizeRange.x = min(sizeRange.x, size.x);
             sizeRange.y = max(sizeRange.y, size.y);
         }