725532 - IonMonkey: Handle JSOP_DEFVAR, JSOP_DEFCONST.

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Sean Stangl [:sstangl]]

Handling JSOP_DEFVAR will halve the execution time of SS-1.0's 3d-morph.

Comment 1

[Sean Stangl [:sstangl]]

Attachment: Refactor DefVarOrConstoperation to be Ion-friendly.

To land against m-i.

Comment 2

[Sean Stangl [:sstangl]]

Attachment: Handle JSOP_DEFVAR and JSOP_DEFCONST (which are basically the same).

We don't handle isEvalFrame() yet -- that changes the |attrs| variable passed to DefVarOrConstOperation(). Notes were left to make sure this isn't accidentally overlooked in the future.

Comment 3

[Sean Stangl [:sstangl]]

http://hg.mozilla.org/integration/mozilla-inbound/rev/3c017b5fd4a5

Comment 4

[Marco Bonardo [:mak]]

https://hg.mozilla.org/mozilla-central/rev/3c017b5fd4a5

Comment 5

[David Anderson [:dvander] - inactive, e-mail if emergency]

Comment on attachment 596845 [details] [diff] [review]
Handle JSOP_DEFVAR and JSOP_DEFCONST (which are basically the same).

Review of attachment 596845 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/src/ion/IonBuilder.cpp
@@ +3623,5 @@
> +    PropertyName *name = script->getName(index);
> +
> +    // Bake in attrs.
> +    uintN attrs = JSPROP_ENUMERATE;
> +    // TODO: if isEvalFrame(), then attrs |= JSPROP_PERMANENT.

change :TODO: to an assert or something (or add a test case that would fail?)

Comment 6

[Sean Stangl [:sstangl]]

http://hg.mozilla.org/projects/ionmonkey/rev/99423ac24643