unilab.ipc.replay_pipelines.transfer.xpu.XpuReplayTransferBackend
-
class unilab.ipc.replay_pipelines.transfer.xpu.XpuReplayTransferBackend[source]
Bases: object
XPU stream/event backend for packed replay batch transfer.
- Parameters:
-
Methods
__init__(*, device, ring_depth)
|
|
allocate_device_slots(*, count, shape, dtype)
|
|
clear_ready(slot)
|
|
close()
|
|
ready_query(slot)
|
|
register_host_slots(slots)
|
|
submit_h2d(*, slot, dst, src, metadata, ...)
|
|
synchronize_ready(slot)
|
|
wait_current_stream_for_ready(slot)
|
|
Attributes
-
device_family = 'xpu'
-
h2d_submitter = 'torch_xpu_copy_stream'
-
host_memory_kind = 'pageable_shared'
-
host_pinned = False
-
direct_pinned_shared = False
-
supports_async_submit = True
-
supports_timing_events = False
-
__init__(*, device, ring_depth)[source]
- Parameters:
-
-
register_host_slots(slots)[source]
- Parameters:
slots (list[Tensor])
- Return type:
None
-
allocate_device_slots(*, count, shape, dtype)[source]
- Parameters:
-
- Return type:
list[Tensor]
-
submit_h2d(*, slot, dst, src, metadata, trace_recorder, trace_cuda_events, h2d_bytes, pack_layout, pack_executor)[source]
- Parameters:
-
- Return type:
float
-
clear_ready(slot)[source]
- Parameters:
slot (int)
- Return type:
None
-
ready_query(slot)[source]
- Parameters:
slot (int)
- Return type:
bool
-
synchronize_ready(slot)[source]
- Parameters:
slot (int)
- Return type:
None
-
wait_current_stream_for_ready(slot)[source]
- Parameters:
slot (int)
- Return type:
None
-
close()[source]
- Return type:
None