wangbo/ComfyUI
1
0
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2026-02-03 10:10:20 +08:00
Code Issues Actions 10 Packages Projects Releases Wiki Activity

disable async pin population

Browse Source
This commit is contained in:
Rattus 2026-01-26 00:53:07 +10:00
parent 12e1560dcc
commit 878c5156d6
1 changed files with 5 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
comfy/ops.py
View File

@ -187,13 +187,11 @@ def cast_bias_weight_with_vbar(s, dtype, device, bias_dtype, non_blocking, compu
if pin is not None:
xfer_dest = comfy.memory_management.interpret_gathered_like([ pin ], xfer_dest)[0]
#FIXME: This might be the wrong thing to do. Some reading suggests the DMA engine
#is posted writes and the compute stream could just fire and forget here. That
#would save this sync and some stalling on the offload stream that is better off
#running ahead to the next layer to read.
if offload_stream is not None:
offload_stream.wait_stream(comfy.model_management.current_stream(device))
comfy.model_management.cast_to(xfer_dest, device=pin.device, non_blocking=non_blocking, stream=offload_stream, r=pin)
#FIXME: put this on nsight and see if its worth offloading to the pin with
#the offload stream. This adds extra sync requirements on xfer_dest in addition to:
#if offload_stream is not None:
# offload_stream.wait_stream(comfy.model_management.current_stream(device))
comfy.model_management.cast_to(xfer_dest, device=pin.device, non_blocking=non_blocking, stream=None, r=pin)
#FIXME: weird offload return protocol
return weight, bias, (offload_stream, device if signature is not None else None, None)