New instruction string.from_code_point

[Liedtke]

Following up on #5 (comment) we decided to implement string.from_code_point for performance evaluations in V8 (see also #58):

• string.from_code_point(codepoint: i32) -> ref string, opcode 0xfba9
• returns a string consisting of the provided codepoint
• the instruction traps if the argument is not a valid unicode codepoint

[gkdn]

Shouldn't this be returning ref string instead?
@Liedtke @jakobkummerow

[gkdn]

BTW, I think also binaryen mismatches Chrome here and assumes the instruction returns ref string which results in code rejected by V8. Adding @kripken to verify.

[Liedtke]

@gkdn Yeah, the returned value is never null. I'll adapt the return type in v8 and in my comment above.

[MaxGraey]

What should happen for invalid code points? Like 0x110000 and above. In JavaScript this leads to an exception. At least for String.fromCodePoint

[Liedtke]

@MaxGraey The instruction traps for invalid codepoints.

[gkdn]

@Liedtke do you have link to the dashboard to track the release? or if you know dev-channel version to get the fix that would work as well.

[jakobkummerow]

@gkdn It's in Dev 112.0.5612, see https://chromiumdash.appspot.com/commit/54b016989406901e819d3090a92739e7bc1c98b9.