diff --git a/arraycontext/impl/pyopencl/__init__.py b/arraycontext/impl/pyopencl/__init__.py
index 84d5f483..36721013 100644
--- a/arraycontext/impl/pyopencl/__init__.py
+++ b/arraycontext/impl/pyopencl/__init__.py
@@ -56,6 +56,26 @@
 
 # {{{ PyOpenCLArrayContext
 
+
+class PyOpenCLProfilingExecutor:
+    def __init__(self, actx, executor):
+        self.executor = executor
+        self.t_unit = executor.t_unit
+        self.actx = actx
+
+    def __call__(self, queue, **kwargs):
+        if self.actx.profile_kernels:
+            import pyopencl as cl
+            start_evt = cl.enqueue_marker(queue)
+
+        evt, result = self.executor(queue, **kwargs)
+
+        if self.actx.profile_kernels:
+            self.actx._add_profiling_events(start_evt, evt, t_unit.entrypoint)
+
+        return evt, result
+
+
 class PyOpenCLArrayContext(ArrayContext):
     """
     A :class:`ArrayContext` that uses :class:`pyopencl.array.Array` instances
@@ -246,7 +266,8 @@ def call_loopy(self, t_unit, **kwargs):
             executor = self._loopy_transform_cache[t_unit]
         except KeyError:
             orig_t_unit = t_unit
-            executor = self.transform_loopy_program(t_unit).executor(self.context)
+            executor = PyOpenCLProfilingExecutor(self,
+                    self.transform_loopy_program(t_unit).executor(self.context))
             self._loopy_transform_cache[orig_t_unit] = executor
             del orig_t_unit