820406 - BaselineCompiler: Store the scope chain

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Jan de Mooij [:jandem]]

Need this to compile DEFVAR, NAME etc.

Comment 1

[Jan de Mooij [:jandem]]

Attachment: Patch

I tried to handle heavyweight functions too, but we have to compile more ops to test it. So this just aborts on heavyweight functions until we support DEFVAR, NAME, LAMBDA etc.

Comment 2

[Jan de Mooij [:jandem]]

Attachment: Patch

Comment 3

[Kannan Vijayan [:djvj]]

Comment on attachment 691368 [details] [diff] [review]
Patch

Review of attachment 691368 [details] [diff] [review]:
-----------------------------------------------------------------

Thanks for taking care of this.  Unblocks some important ops for implementation.

::: js/src/ion/arm/IonFrames-arm.h
@@ +462,5 @@
>      uint32_t loScratchValue;
>      uint32_t hiScratchValue;
>      size_t frameSize;
> +    JSObject *scopeChain;
> +    uint32_t dummy; // Keep frame 8-byte aligned.

I don't understand the need for this.  Without it, BaselineFrame is 16 bytes on ARM.  This bumps it to 20 bytes.  Are we expecting the stack to be at a 4-byte offset when this is pushed?

Comment 4

[Jan de Mooij [:jandem]]

(In reply to Kannan Vijayan [:djvj] from comment #3)
>
> I don't understand the need for this.  Without it, BaselineFrame is 16 bytes
> on ARM.  This bumps it to 20 bytes.  Are we expecting the stack to be at a
> 4-byte offset when this is pushed?

Yeah, the 4-byte offset is for the old frame pointer we push on the stack. See the BaselineFrame static assert, it asserts (sizeof(BaselineFrame) + FramePointerOffset) % 8 == 0.

Comment 5

[Jan de Mooij [:jandem]]

https://hg.mozilla.org/projects/ionmonkey/rev/bc21eceb1c25