Add, use enqueue_fill

Author: inducer

doc/runtime_memory.rst

@@ -281,6 +281,8 @@ Transfers
 
 .. autofunction:: enqueue_copy(queue, dest, src, **kwargs)
 
+.. autofunction:: enqueue_fill(queue, dest, src, **kwargs)
+
 Mapping Memory into Host Address Space
 --------------------------------------
 

pyopencl/__init__.py

@@ -22,6 +22,7 @@
 
 from sys import intern
 from warnings import warn
+from typing import Union, Any
 
 from pyopencl.version import VERSION, VERSION_STATUS, VERSION_TEXT  # noqa
 
@@ -1815,6 +1816,26 @@ def enqueue_copy(queue, dest, src, **kwargs):
 # }}}
 
 
+# {{{ enqueue_fill
+
+def enqueue_fill(queue: CommandQueue, dest: Union[MemoryObjectHolder, SVM],
+        pattern: Any, size: int, *, offset: int = 0, wait_for=None) -> Event:
+    """
+    .. versionadded:: 2022.2
+    """
+    if isinstance(dest, MemoryObjectHolder):
+        return enqueue_fill_buffer(queue, dest, pattern, offset, size, wait_for)
+    elif isinstance(dest, SVM):
+        if offset:
+            raise NotImplementedError("enqueue_fill with SVM does not yet support "
+                    "offsets")
+        return enqueue_svm_memfill(queue, dest, pattern, size, wait_for)
+    else:
+        raise TypeError(f"enqueue_fill does not know how to fill '{type(dest)}'")
+
+# }}}
+
+
 # {{{ image creation
 
 DTYPE_TO_CHANNEL_TYPE = {

pyopencl/array.py

@@ -1458,8 +1458,8 @@ def _zero_fill(self, queue=None, wait_for=None):
         # https://github.com/inducer/pyopencl/issues/395
         if cl_version_gtr_1_2 and not (on_nvidia and self.nbytes >= 2**31):
             self.add_event(
-                    cl.enqueue_fill_buffer(queue, self.base_data, np.int8(0),
-                        self.offset, self.nbytes, wait_for=wait_for))
+                    cl.enqueue_fill(queue, self.base_data, np.int8(0),
+                        self.nbytes, offset=self.offset, wait_for=wait_for))
         else:
             zero = np.zeros((), self.dtype)
             self.fill(zero, queue=queue)