diff --git a/doc/misc.rst b/doc/misc.rst
index 9043cf526abb221f3c4c5e51c8d92f4a8502a58f..7069a5844e7818b1294d3fef5860b31ab271942c 100644
--- a/doc/misc.rst
+++ b/doc/misc.rst
@@ -41,7 +41,7 @@ On Linux, type:
 #.  ``conda install pocl``
 
 to install a CPU-based OpenCL driver on Linux. On Windows, you may install e.g.
-the `OpenCL driver from Intel <https://software.intel.com/en-us/articles/opencl-drivers#latest_CPU_runtime>`.
+the `OpenCL driver from Intel <https://software.intel.com/en-us/articles/opencl-drivers#latest_CPU_runtime>`_.
 OS X has support for OpenCL built into the operating system and does not need
 additional software to run code based on PyOpenCL (but see below).
 
diff --git a/examples/demo.py b/examples/demo.py
index 62c0f7ee5fe975f0d4097be41396b558e82eec50..faac547ee7283ab01873549c604655fa9ac2b6d0 100644
--- a/examples/demo.py
+++ b/examples/demo.py
@@ -12,8 +12,11 @@ ctx = cl.create_some_context()
 queue = cl.CommandQueue(ctx)
 
 mf = cl.mem_flags
-a_g = cl.Buffer(ctx, mf.READ_ONLY | mf.COPY_HOST_PTR, hostbuf=a_np)
-b_g = cl.Buffer(ctx, mf.READ_ONLY | mf.COPY_HOST_PTR, hostbuf=b_np)
+a_g = cl.Buffer(ctx, mf.READ_WRITE, a_np.nbytes)
+b_g = cl.Buffer(ctx, mf.READ_WRITE, b_np.nbytes)
+
+cl.enqueue_copy(queue, a_g, a_np)
+cl.enqueue_copy(queue, b_g, b_np)
 
 prg = cl.Program(ctx, """
 __kernel void sum(
@@ -25,7 +28,7 @@ __kernel void sum(
 """).build()
 
 res_g = cl.Buffer(ctx, mf.WRITE_ONLY, a_np.nbytes)
-prg.sum(queue, a_np.shape, None, a_g, b_g, res_g)
+prg.sum(queue, (500, 10), (16, 16), a_g, b_g, res_g, g_times_l=True)
 
 res_np = np.empty_like(a_np)
 cl.enqueue_copy(queue, res_np, res_g)