Merge pull request #75 from davidweichiang/master

slice size should round up, not down

Author: inducer

pycuda/gpuarray.py

@@ -786,7 +786,7 @@ def __getitem__(self, index):
 
                 array_stride = self.strides[array_axis]
 
-                new_shape.append((stop-start)//idx_stride)
+                new_shape.append((stop-start-1)//idx_stride+1)
                 new_strides.append(idx_stride*array_stride)
                 new_offset += array_stride*start