mirror of
https://github.com/pytorch/pytorch.git
synced 2025-10-20 21:14:14 +08:00
b8550f527f
# Motivation Support GPU trace on XPU backend. Add GPU trace to xpu runtime. It is beneficial to generalize the device caching allocator in the next step. Pull Request resolved: https://github.com/pytorch/pytorch/pull/121795 Approved by: https://github.com/EikanWang, https://github.com/gujinghui, https://github.com/jgong5, https://github.com/albanD ghstack dependencies: #121794
76 lines
2.3 KiB
Python
76 lines
2.3 KiB
Python
from typing import Callable
|
|
|
|
from torch._utils import CallbackRegistry
|
|
|
|
|
|
EventCreationCallbacks: "CallbackRegistry[int]" = CallbackRegistry(
|
|
"XPU event creation"
|
|
)
|
|
EventDeletionCallbacks: "CallbackRegistry[int]" = CallbackRegistry(
|
|
"XPU event deletion"
|
|
)
|
|
EventRecordCallbacks: "CallbackRegistry[int, int]" = CallbackRegistry(
|
|
"XPU event record"
|
|
)
|
|
EventWaitCallbacks: "CallbackRegistry[int, int]" = CallbackRegistry(
|
|
"XPU event wait"
|
|
)
|
|
MemoryAllocationCallbacks: "CallbackRegistry[int]" = CallbackRegistry(
|
|
"XPU memory allocation"
|
|
)
|
|
MemoryDeallocationCallbacks: "CallbackRegistry[int]" = CallbackRegistry(
|
|
"XPU memory deallocation"
|
|
)
|
|
StreamCreationCallbacks: "CallbackRegistry[int]" = CallbackRegistry(
|
|
"XPU stream creation"
|
|
)
|
|
DeviceSynchronizationCallbacks: "CallbackRegistry[[]]" = CallbackRegistry(
|
|
"XPU device synchronization"
|
|
)
|
|
StreamSynchronizationCallbacks: "CallbackRegistry[int]" = CallbackRegistry(
|
|
"XPU stream synchronization"
|
|
)
|
|
EventSynchronizationCallbacks: "CallbackRegistry[int]" = CallbackRegistry(
|
|
"XPU event synchronization"
|
|
)
|
|
|
|
|
|
def register_callback_for_event_creation(cb: Callable[[int], None]) -> None:
|
|
EventCreationCallbacks.add_callback(cb)
|
|
|
|
|
|
def register_callback_for_event_deletion(cb: Callable[[int], None]) -> None:
|
|
EventDeletionCallbacks.add_callback(cb)
|
|
|
|
|
|
def register_callback_for_event_record(cb: Callable[[int, int], None]) -> None:
|
|
EventRecordCallbacks.add_callback(cb)
|
|
|
|
|
|
def register_callback_for_event_wait(cb: Callable[[int, int], None]) -> None:
|
|
EventWaitCallbacks.add_callback(cb)
|
|
|
|
|
|
def register_callback_for_memory_allocation(cb: Callable[[int], None]) -> None:
|
|
MemoryAllocationCallbacks.add_callback(cb)
|
|
|
|
|
|
def register_callback_for_memory_deallocation(cb: Callable[[int], None]) -> None:
|
|
MemoryDeallocationCallbacks.add_callback(cb)
|
|
|
|
|
|
def register_callback_for_stream_creation(cb: Callable[[int], None]) -> None:
|
|
StreamCreationCallbacks.add_callback(cb)
|
|
|
|
|
|
def register_callback_for_device_synchronization(cb: Callable[[], None]) -> None:
|
|
DeviceSynchronizationCallbacks.add_callback(cb)
|
|
|
|
|
|
def register_callback_for_stream_synchronization(cb: Callable[[int], None]) -> None:
|
|
StreamSynchronizationCallbacks.add_callback(cb)
|
|
|
|
|
|
def register_callback_for_event_synchronization(cb: Callable[[int], None]) -> None:
|
|
EventSynchronizationCallbacks.add_callback(cb)
|