[core][move semantics][3/n]Populate task args with is_move for movable refs by karticam · Pull Request #66803 · ray-project/ray

karticam · 2026-10-07T02:02:14Z

This is PR 3 of a series of PRs to implement move semantics for plasma objects in core.

The complete design doc is here: https://docs.google.com/document/d/1ZJhgRmwVyKfNCR21CWf6xMcc0B6HJwSPQsfebTRvnnE/edit?tab=t.0#heading=h.7wcod1f9afpg

This marks each consume-once argument of an actor task with a new TaskArg.is_move flag in the task spec. After #66767, passing a consume-once ref to an actor task changes its state to MOVED; this PR records the same fact in the task spec, so that the consumer's raylet can later pull the object as a move instead of a copy. Nothing reads the flag yet: the consumer side comes in the next PRs.

Changes

New proto field TaskArg.is_move. True when the task consumes the argument, which was created with _consume_once=True. Only set on pass-by-ref arguments. Arguments that aren't consume-once add nothing on the wire.
ReferenceCounter::TryCommitMoves returns which args it moved. Its return type changes from Status to StatusOr<std::vector<bool>>: on success, one entry per arg ID, true if this call moved the object (it was MOVABLE). The entries are filled in the validation pass, so both entries of a repeated arg (c.remote(r, r), allowed by [core][move semantics][2/n] Change state to moved on consumption and add enforcements #66767) are true. NOT_MOVABLE args are false. On rejection (an arg already MOVED) it returns the same InvalidArgument as before.
SubmitActorTask sets the flag. The commit loop also records each by-ref arg's index. After the task spec is built, the moved indices get TaskSpecBuilder::SetArgIsMove(i) (new). The proto arg order is the args order, so the indices line up.
New accessor TaskSpecification::ArgIsMove(i), for the consumer side in the next PRs.
The flag is cleared when the arg is inlined. A small consume-once object lives in the caller's in-memory store, and the dependency resolver copies its value into the task spec (InlineDependencies). There's nothing to pull or pin, so InlineDependencies clears is_move. An object in plasma stays by-ref and keeps the flag. (Tensor-transport args, which keep their object_ref when inlined, can't be consume-once: [core][move semantics][1/n] Adding consume_once flag in actor methods #66620 rejects _consume_once with tensor_transport.)
Task retries reuse the stored task spec, so a retried task keeps is_move.

_{Stack created with GitHub Stacks CLI • Give Feedback 💬}

gemini-code-assist · 2026-10-07T02:02:18Z

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

cursor

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 913a17f. Configure here.}

cursor · 2026-10-08T20:37:26Z


+  /// Return true if this task consumes the argument at the given index, which was
+  /// created with consume_once.
+  bool ArgIsMove(size_t arg_index) const;


New APIs missing Doxygen tags

Low Severity

⚠️ Document functions and classes with Doxygen /** ... */ block comments using @ tags (@brief, and @param/@return where applicable). ArgIsMove and SetArgIsMove have brief comments but omit @param/@return for arg_index and the return value.

Additional Locations (1)

src/ray/common/task/task_util.h#L253-L258

^{Triggered by project rule: Bugbot Rules}

^{Reviewed by Cursor Bugbot for commit 913a17f. Configure here.}

gemini-code-assist

Code Review

This pull request introduces tracking for arguments that are consumed (moved) by a task (created with consume_once). It adds an is_move field to the TaskArg protobuf message, updates ReferenceCounter::TryCommitMoves to return a list of booleans indicating which arguments were moved, and ensures that is_move is cleared if an argument is inlined during dependency resolution. It also includes corresponding unit tests and updates to task building and submission. I have no feedback to provide as there are no review comments.

Sparks0219 · 2026-10-30T23:18:03Z

  ASSERT_TRUE(dependencies_resolved.get_future().get());
  ASSERT_TRUE(ok);
  ASSERT_TRUE(task.ArgByRef(0));
+  // An argument that stays in plasma keeps is_move.


remove comment

Sparks0219 · 2026-10-30T23:18:10Z

  ASSERT_FALSE(task.ArgByRef(1));
  ASSERT_NE(task.ArgData(0), nullptr);
  ASSERT_NE(task.ArgData(1), nullptr);
+  // An inlined argument is not moved.


remove comment

Sparks0219 · 2026-10-30T23:24:08Z

          }

          mutable_arg->set_is_inlined(true);
+          // The value is now in the task spec (inlined), so there is nothing to move.


We should document this, for in-lined objects it's a no-op

Sparks0219 · 2026-10-30T23:34:48Z

                       move_state);
    return id;
  };
  ObjectID plain = add_owned_object(MoveState::NOT_MOVABLE);


can you rename plain to not_movable

Sparks0219 · 2026-10-30T23:36:15Z


-  // A repeated ID is moved once; NOT_MOVABLE and unknown IDs are left alone.
-  ASSERT_TRUE(rc->TryCommitMoves({plain, movable_1, movable_1, unknown}).ok());
+  // A repeated ID is moved once and both entries are reported as moves; NOT_MOVABLE


can you clarify the state machine for MOVABLE here, would help

Signed-off-by: kartica.modi <kartica.modi@anyscale.com>

karticam added this pull request to stack #66768 October 7, 2026 02:02

karticam force-pushed the karticam/is-move-in-task-args branch from 6b6c4f0 to 6ec8cbb Compare October 7, 2026 02:02

karticam marked this pull request as ready for review October 8, 2026 20:33

karticam requested a review from a team as a code owner October 8, 2026 20:33

cursor Bot reviewed Oct 8, 2026

View reviewed changes

gemini-code-assist Bot reviewed Oct 8, 2026

View reviewed changes

karticam assigned Sparks0219 Oct 8, 2026

ray-gardener Bot added the core Issues that should be addressed in Ray Core label Oct 9, 2026

karticam mentioned this pull request Oct 15, 2026

[core][move semantics][4/n] Consumer raylet records which pulls are moves #67041

Open

karticam added go add ONLY when ready to merge, run all tests unstale A PR that has been marked unstale. It will not get marked stale again if this label is on it. labels Oct 16, 2026

karticam force-pushed the karticam/is-move-in-task-args branch from 913a17f to d4703ac Compare October 21, 2026 05:14

Sparks0219 reviewed Oct 30, 2026

View reviewed changes

karticam mentioned this pull request Nov 3, 2026

[core][move semantics][1/n] Adding consume_once flag in actor methods #66620

Merged

karticam force-pushed the karticam/is-move-in-task-args branch 2 times, most recently from cbe3351 to 95c6f8b Compare November 3, 2026 03:07

karticam added 2 commits November 14, 2026 02:05

Populate task args with is_move for movable refs

7c762f9

Signed-off-by: kartica.modi <kartica.modi@anyscale.com>

Adding comment

2ae4e61

Signed-off-by: kartica.modi <kartica.modi@anyscale.com>

karticam force-pushed the karticam/is-move-in-task-args branch from 95c6f8b to 2ae4e61 Compare November 14, 2026 02:36

Navigation Menu

Repository navigation

[core][move semantics][3/n]Populate task args with is_move for movable refs - #66803

[core][move semantics][3/n]Populate task args with is_move for movable refs#66803
karticam wants to merge 2 commits into
karticam/move-sem-consume-and-enforcementsfrom
karticam/is-move-in-task-args

karticam commented Oct 7, 2026 •

edited

Loading

Uh oh!

gemini-code-assist Bot commented Oct 7, 2026

Uh oh!

cursor Bot left a comment

Uh oh!

Uh oh!

cursor Bot Oct 8, 2026

Uh oh!

gemini-code-assist Bot left a comment

Uh oh!

Sparks0219 Oct 30, 2026

Uh oh!

Sparks0219 Oct 30, 2026

Uh oh!

Sparks0219 Oct 30, 2026

Uh oh!

Sparks0219 Oct 30, 2026

Uh oh!

Sparks0219 Oct 30, 2026

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

karticam commented Oct 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changes

Uh oh!

gemini-code-assist Bot commented Oct 7, 2026

Uh oh!

cursor Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

cursor Bot Oct 8, 2026

Choose a reason for hiding this comment

New APIs missing Doxygen tags

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

Sparks0219 Oct 30, 2026

Choose a reason for hiding this comment

Uh oh!

Sparks0219 Oct 30, 2026

Choose a reason for hiding this comment

Uh oh!

Sparks0219 Oct 30, 2026

Choose a reason for hiding this comment

Uh oh!

Sparks0219 Oct 30, 2026

Choose a reason for hiding this comment

Uh oh!

Sparks0219 Oct 30, 2026

Choose a reason for hiding this comment

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

karticam commented Oct 7, 2026 •

edited

Loading