Skip to content

Commit

Permalink
Merge pull request #23924 from jburnim:jburnim_doc_typo_fix
Browse files Browse the repository at this point in the history
PiperOrigin-RevId: 678896733
  • Loading branch information
Google-ML-Automation committed Sep 25, 2024
2 parents e4ca4f5 + a1f2edc commit 911acf1
Showing 1 changed file with 2 additions and 2 deletions.
4 changes: 2 additions & 2 deletions docs/pallas/async_note.md
Original file line number Diff line number Diff line change
Expand Up @@ -37,7 +37,7 @@ Now imagine we aren’t using XLA’s `ppermute` but have our own custom Pallas
```py
def ppermute_kernel(x_ref, y_ref, send_sem, recv_sem):
right_neighbor = ...
descriptor = pltpu.make_remote_async_copy(x_ref, y_ref, send_sem, recv_sem, device_id=right_neighbor)
descriptor = pltpu.make_async_remote_copy(x_ref, y_ref, send_sem, recv_sem, device_id=right_neighbor)
descriptor.start()
descriptor.wait_send()
descriptor.wait_recv()
Expand All @@ -54,7 +54,7 @@ def add_one(x_ref, z_ref):

def ppermute_add_one_kernel(x_ref, y_ref, z_ref, send_sem, recv_sem):
right_neighbor = ...
descriptor = pltpu.make_remote_async_copy(x_ref, y_ref, send_sem, recv_sem, device_id=right_neighbor)
descriptor = pltpu.make_async_remote_copy(x_ref, y_ref, send_sem, recv_sem, device_id=right_neighbor)
descriptor.start()

# Explicitly schedule inner kernel between start/wait
Expand Down

0 comments on commit 911acf1

Please sign in to comment.