Skip to content
Draft
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 11 additions & 2 deletions pytato/distributed/execute.py
Original file line number Diff line number Diff line change
Expand Up @@ -110,7 +110,8 @@ def execute_distributed_partition(
queue: Any, mpi_communicator: Any,
*,
allocator: Any | None = None,
input_args: dict[str, Any] | None = None) -> dict[str, Any]:
input_args: dict[str, Any] | None = None,
actx: Any | None = None) -> dict[str, Any]:

if input_args is None:
input_args = {}
Expand Down Expand Up @@ -168,10 +169,18 @@ def _get_partition_input_name_refcount(partition: DistributedGraphPartition) \
def exec_ready_part(part: DistributedGraphPart) -> None:
inputs = {k: context[k] for k in part.all_input_names()}

_evt, result_dict = prg_per_partition[part.pid](queue,
if actx and actx.profile_kernels:
import pyopencl as cl
start_evt = cl.enqueue_marker(queue)

evt, result_dict = prg_per_partition[part.pid](queue,
allocator=allocator,
**inputs)

if actx and actx.profile_kernels:
name = next(iter(prg_per_partition[part.pid].program.entrypoints))
actx._add_profiling_events(start_evt, evt, name)

context.update(result_dict)

for name, send_nodes in part.name_to_send_nodes.items():
Expand Down
Loading