1003918 - Rename ionTop to jitTop

Status: RESOLVED FIXED

(Core :: JavaScript Engine: JIT, defect)

Description

[Jan de Mooij [:jandem]]

Attachment: Patch

This week's cleanup patch. There were more uses of ionTop than I expected.

This patch also moves ionReturnOverride_ from JSRuntime to JitRuntime.

Comment 1

[Shu-yu Guo [:shu]]

Comment on attachment 8415372 [details] [diff] [review]
Patch

Review of attachment 8415372 [details] [diff] [review]:
-----------------------------------------------------------------

Looks good to me!

::: js/src/jit/JitCompartment.h
@@ +356,5 @@
>          return forkJoinGetSliceStub_;
>      }
> +
> +    bool hasIonReturnOverride() const {
> +        return !ionReturnOverride_.isMagic();

Just wondering: can we tighten this to !isMagic(JS_ARG_POSION), or are other magic values supported as return values?

::: js/src/jit/x64/MacroAssembler-x64.h
@@ +1300,5 @@
>          orq(Imm32(type), frameSizeReg);
>      }
>  
>      // Save an exit frame (which must be aligned to the stack pointer) to
> +    // ThreadData::jitTop of the main thread.

Nit: preexisting; ThreadData -> PerThreadData

::: js/src/jit/x86/MacroAssembler-x86.h
@@ +1097,5 @@
>          orl(Imm32(type), frameSizeReg);
>      }
>  
>      // Save an exit frame (which must be aligned to the stack pointer) to
> +    // ThreadData::jitTop of the main thread.

Ditto.

Comment 2

[Jan de Mooij [:jandem]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/fa468cb36e5e

(In reply to Shu-yu Guo [:shu] from comment #1)
> Just wondering: can we tighten this to !isMagic(JS_ARG_POSION), or are other
> magic values supported as return values?

Good idea, done. We shouldn't see other magic values there AFAIK.

Comment 3

[Jan de Mooij [:jandem]]

Followup to unbreak non-unified builds:

https://hg.mozilla.org/integration/mozilla-inbound/rev/7210c4f043a5

Comment 4

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/fa468cb36e5e
https://hg.mozilla.org/mozilla-central/rev/7210c4f043a5