1061229 - SIMD (interpreter): Add coercive calls

Status: RESOLVED FIXED

(Core :: JavaScript: Standard Library, defect)

Description

[Benjamin Bouvier [:bbouvier] (inactive)]

Coercive calls (with the form x=SIMD.int32x4(x)) shouldn't be only a part of asm.js, but should be implemented in the interpreter as well. That will also allow us to compare asm.js and non-asmjs (interpreter only) variants of the demos.

Comment 1

[Benjamin Bouvier [:bbouvier] (inactive)]

Attachment: bug1061229.patch

Steve, mrgiggles told you can review SIMD.cpp :) That should be a really easy review, and hopefully we'd land this before the end of the week.

Comment 2

[Benjamin Bouvier [:bbouvier] (inactive)]

Attachment: bug1061229.patch

Hannes proposed to review this one :)

Comment 3

[Hannes Verschore [:h4writer]]

Comment on attachment 8482614 [details] [diff] [review]
bug1061229.patch

Review of attachment 8482614 [details] [diff] [review]:
-----------------------------------------------------------------

Lgtm

::: js/src/builtin/SIMD.cpp
@@ +307,5 @@
>      const unsigned LANES = 4;
>  
> +    if (args.length() == 1) {
> +        // X4 type used as a coercion
> +        if (!CheckVectorObject(args[0], args.callee().as<X4TypeDescr>().type())) {

You can move the declaration of descr[0] above here. So you can just use "descr.type()" instead.

[0] http://dxr.mozilla.org/mozilla-central/source/js/src/builtin/SIMD.cpp?from=simd.cpp&case=true#315"

Comment 4

[Benjamin Bouvier [:bbouvier] (inactive)]

Good call!

https://hg.mozilla.org/integration/mozilla-inbound/rev/2cd12da0c4e6

Comment 5

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/2cd12da0c4e6