Skip to content

test(cuda.core): check the event-wait target with device timestamps #3262

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Andy-Jost merged 2 commits into from
Oct 30, 2026
+20 −28
Merged

test(cuda.core): check the event-wait target with device timestamps #3262

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
2803d30
test(cuda.core): check the event-wait target with device timestamps
saiyambharara Oct 28, 2026
684c606
test(cuda.core): skip the event-wait timing check below sm70
saiyambharara Oct 29, 2026
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
48 changes: 20 additions & 28 deletions cuda_core/tests/graph/test_graph_node_update.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,8 +13,9 @@

import pytest
from helpers.graph_kernels import compile_common_kernels
from helpers.nanosleep_kernel import NanosleepKernel

from cuda.core import Device, DeviceMemoryResource, LaunchConfig, LegacyPinnedMemoryResource
from cuda.core import Device, DeviceMemoryResource, EventOptions, LaunchConfig, LegacyPinnedMemoryResource
from cuda.core._memory._synchronous_memory_resource import _SynchronousMemoryResource
from cuda.core._utils._weak_handles import weak_handle
from cuda.core._utils.cuda_utils import CUDAError, driver, handle_return
Expand Down Expand Up @@ -142,42 +143,33 @@ def assert_exec_uses(graph, expected):


def _event_wait_case(device):
"""Keep the selected event pending to identify each exec's wait target."""
original = device.create_event()
replacement = device.create_event()
invalid_replacement = device.create_event()
"""Identify each exec's wait target from device timestamps."""
event_options = EventOptions(timing_enabled=True)
original = device.create_event(options=event_options)
replacement = device.create_event(options=event_options)
invalid_replacement = device.create_event(options=event_options)
invalid_replacement.close()
after_wait = device.create_event(options=event_options)

callback_called = threading.Event()
graph_def = GraphDefinition()
node = graph_def.wait(original)
node.callback(callback_called.set)
node.record(after_wait)

def assert_exec_uses(graph, expected):
producer_started = threading.Event()
producer_release = threading.Event()

def blocking_callback():
producer_started.set()
producer_release.wait(timeout=30)

producer_def = GraphDefinition()
producer_def.callback(blocking_callback).record(expected)
producer_graph = producer_def.instantiate()
if device.compute_capability.major < 7:
pytest.skip("__nanosleep is only available starting Volta (sm70)")
Comment on lines +159 to +160

@coderabbitai coderabbitai Bot Oct 29, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Minor · Performance & Scalability · Quick win

important: Avoid recompiling NanosleepKernel for every assertion.

The guard now runs inside assert_exec_uses, so NanosleepKernel(device) compiles once per execution check instead of once per case. This helper is called multiple times for the event-wait case. Move the capability check before case setup, then construct and reuse delay once. This also skips unsupported devices before creating case resources. (github.com)

delay = NanosleepKernel(device)
producer_stream = device.create_stream()
consumer_stream = device.create_stream()

callback_called.clear()
producer_graph.launch(producer_stream)
try:
assert producer_started.wait(timeout=5)
graph.launch(consumer_stream)
assert not callback_called.wait(timeout=0.1)
finally:
producer_release.set()
producer_stream.sync()
consumer_stream.sync()
assert callback_called.is_set()
delay.launch(producer_stream)
producer_stream.record(expected)
graph.launch(consumer_stream)
producer_stream.sync()
consumer_stream.sync()

# Waiting on any other event lets after_wait land during the delay.
assert after_wait - expected >= 0

return _DefinitionUpdateCase(
graph_def=graph_def,
Expand Down
Loading