Update array.copy semantics to handle packed types by tlively · Pull Request #416 · WebAssembly/gc

tlively · 2023-08-24T19:18:56Z

Previously the array.copy semantics were defined in terms of executing array.get
on the source array, which is not correct if the source array contains a packed
type. Update the semantics prose and formalism to choose between array.get and
array.get_u depending on whether the array contents are packed. The interpreter
does not need updating because it already handled the difference correctly.

Previously the array.copy semantics were defined in terms of executing array.get on the source array, which is not correct if the source array contains a packed type. Update the semantics prose and formalism to choose between array.get and array.get_u depending on whether the array contents are packed. The interpreter does not need updating because it already handled the difference correctly.

tlively · 2023-08-24T19:19:08Z

Current dependencies on/for this PR:

main
- PR Update array.copy semantics to handle packed types #416 👈

This comment was auto-generated by Graphite.

tlively · 2023-08-24T19:26:38Z

+                  \land & (\X{get} = \ARRAYGET~y \vee \X{get} = \ARRAYGETU~y) \\
+                  \land & (F.\AMODULE.\MITYPES[x] = \TARRAY~\mut~\packedtype \Longleftrightarrow \X{get} = \ARRAYGETU~y))


Can you think of a better way to express this? Is there a better symbol to use for bidirectional implication than \Longleftrightarrow?

The symbol would be correct. But I would suggest to factor out the choice of get into an auxiliary function which does a case distinction on the storage type, i.e., something like

getfield(valtype) = array.get getfield(packedtype) = array.get_u

Thanks, I've written this up now. PTAL!

rossberg · 2023-09-06T14:51:31Z

 ......................

-.. todo:: Handle packed fields correctly via array.get_u instead of array.get
+1. Assert: due to :ref:`validation <valid-array.init_data>`, the :ref:`defined type <syntax-deftype>` :math:`F.\AMODULE.\MITYPES[y]` exists.


Suggested change

1. Assert: due to :ref:`validation <valid-array.init_data>`, the :ref:`defined type <syntax-deftype>` :math:`F.\AMODULE.\MITYPES[y]` exists.

1. Assert: due to :ref:`validation <valid-array.copy>`, the :ref:`defined type <syntax-deftype>` :math:`F.\AMODULE.\MITYPES[y]` exists.

rossberg · 2023-09-06T14:52:01Z

+2. Let :math:`\deftype` be the :ref:`defined type <syntax-deftype>` :math:`F.\AMODULE.\MITYPES[y]`.

-2. Pop the value :math:`n` from the stack.
+3. Assert: due to :ref:`validation <valid-array.init_data>`, the :ref:`expansion <aux-expand-deftype>` of :math:`\deftype` is an :ref:`array type <syntax-arraytype>`.


Suggested change

3. Assert: due to :ref:`validation <valid-array.init_data>`, the :ref:`expansion <aux-expand-deftype>` of :math:`\deftype` is an :ref:`array type <syntax-arraytype>`.

3. Assert: due to :ref:`validation <valid-array.copy>`, the :ref:`expansion <aux-expand-deftype>` of :math:`\deftype` is an :ref:`array type <syntax-arraytype>`.

rossberg · 2023-09-06T14:57:36Z

+                  \land & (\X{get} = \ARRAYGET~y \vee \X{get} = \ARRAYGETU~y) \\
+                  \land & (F.\AMODULE.\MITYPES[x] = \TARRAY~\mut~\packedtype \Longleftrightarrow \X{get} = \ARRAYGETU~y))


The symbol would be correct. But I would suggest to factor out the choice of get into an auxiliary function which does a case distinction on the storage type, i.e., something like

getfield(valtype) = array.get getfield(packedtype) = array.get_u

rossberg · 2023-09-08T23:52:57Z

+5. Assert: due to :ref:`validation <valid-array.copy>`, a :ref:`value <syntax-val>` of :ref:`type <syntax-valtype>` :math:`\I32` is on the top of the stack.

-   a. Let :math:`\X{get}` be the instruction :math:`\ARRAYGET~y`.
+6. Pop the value :math:`n` from the stack.


Suggested change

6. Pop the value :math:`n` from the stack.

6. Pop the value :math:`\I32.\CONST~n` from the stack.

rossberg · 2023-09-08T23:53:51Z

-8. Pop the value :math:`n` from the stack.
-
-9. Assert: due to :ref:`validation <valid-array.copy>`, a :ref:`value <syntax-val>` of :ref:`type <syntax-valtype>` :math:`\I32` is on the top of the stack.
+8. Pop the value :math:`s` from the stack.


Suggested change

8. Pop the value :math:`s` from the stack.

8. Pop the value :math:`\I32.\CONST~s` from the stack.

rossberg · 2023-09-08T23:54:25Z

+11. Assert: due to :ref:`validation <valid-array.copy>`, a :ref:`value <syntax-val>` of :ref:`type <syntax-valtype>` :math:`\I32` is on the top of the stack.

-13. Assert: due to :ref:`validation <valid-array.copy>`, a :ref:`value <syntax-val>` of :ref:`type <syntax-valtype>` :math:`\I32` is on the top of the stack.
+12. Pop the value :math:`d` from the stack.


Suggested change

12. Pop the value :math:`d` from the stack.

12. Pop the value :math:`(\I32.\CONST~d)` from the stack.

rossberg · 2023-09-08T23:56:43Z

+   \begin{array}{lll}
+   \getfield(\valtype) &=& \ARRAYGET \\
+   \getfield(\packedtype) &=& \ARRAYGETU \\
+   \end{array}


I'd move this next to the rules (as a "where:" sort of thing), since it is local helper for that specific rule only.

tlively · 2023-09-09T00:48:14Z

Comments addressed. I also updated the formatting of the side conditions now that they are shorter.

rossberg

Thanks!

tlively requested a review from rossberg August 24, 2023 19:18

tlively mentioned this pull request Aug 24, 2023

Spec authoring tracking issue #376

Open

53 tasks

tlively commented Aug 24, 2023

View reviewed changes

rossberg reviewed Sep 6, 2023

View reviewed changes

getfield meta function

8e3581d

rossberg reviewed Sep 8, 2023

View reviewed changes

tlively added 3 commits September 8, 2023 17:35

move auxiliary function

e04ecbd

pop i32.const fixes

31444ad

format conditions on single line

aa173f6

rossberg approved these changes Sep 9, 2023

View reviewed changes

rossberg merged commit d6a2c16 into main Sep 9, 2023

rossberg deleted the array-copy-getu branch September 9, 2023 16:03

		\land & (\X{get} = \ARRAYGET~y \vee \X{get} = \ARRAYGETU~y) \\
		\land & (F.\AMODULE.\MITYPES[x] = \TARRAY~\mut~\packedtype \Longleftrightarrow \X{get} = \ARRAYGETU~y))

	1. Assert: due to :ref:`validation <valid-array.init_data>`, the :ref:`defined type <syntax-deftype>` :math:`F.\AMODULE.\MITYPES[y]` exists.
	1. Assert: due to :ref:`validation <valid-array.copy>`, the :ref:`defined type <syntax-deftype>` :math:`F.\AMODULE.\MITYPES[y]` exists.

	3. Assert: due to :ref:`validation <valid-array.init_data>`, the :ref:`expansion <aux-expand-deftype>` of :math:`\deftype` is an :ref:`array type <syntax-arraytype>`.
	3. Assert: due to :ref:`validation <valid-array.copy>`, the :ref:`expansion <aux-expand-deftype>` of :math:`\deftype` is an :ref:`array type <syntax-arraytype>`.

	6. Pop the value :math:`n` from the stack.
	6. Pop the value :math:`\I32.\CONST~n` from the stack.

	8. Pop the value :math:`s` from the stack.
	8. Pop the value :math:`\I32.\CONST~s` from the stack.

	12. Pop the value :math:`d` from the stack.
	12. Pop the value :math:`(\I32.\CONST~d)` from the stack.

Conversation

tlively commented Aug 24, 2023

Uh oh!

tlively commented Aug 24, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

tlively commented Sep 9, 2023

Uh oh!

rossberg left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

tlively commented Aug 24, 2023 •

edited

Loading