763599 - IonMonkey: Inline string[int32]

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Jan de Mooij [:jandem]]

Attachment: Patch

Pretty easy with MCharCodeAt/MFromCharCode. Sunspider base64 and I think unpack-code do this and it's pretty common on the web.

Comment 1

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

Comment on attachment 631968 [details] [diff] [review]
Patch

Review of attachment 631968 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/src/ion/TypeOracle.cpp
@@ +319,5 @@
> +
> +    if (value->getKnownTypeTag(cx) != JSVAL_TYPE_STRING)
> +        return false;
> +
> +    if (id->getKnownTypeTag(cx) != JSVAL_TYPE_INT32)

Out of curiosity, do calls like this add constraints even though we don't need them in the false case?

Comment 2

[Jan de Mooij [:jandem]]

http://hg.mozilla.org/projects/ionmonkey/rev/17e811e71344

(In reply to David Anderson [:dvander] from comment #1)
> > +
> > +    if (id->getKnownTypeTag(cx) != JSVAL_TYPE_INT32)
> 
> Out of curiosity, do calls like this add constraints even though we don't
> need them in the false case?

Yeah if the type set is empty or has a single type. Brian would it make sense to add a function like this:

if (!id->hasKnownTypeTag(cx, JSVAL_TYPE_INT32))
    return false;

Where hasKnownTypeTag only adds a constraint if either 1) the type tag matches the argument or 2) the type set is empty (to enable this optimization later).