1448965 - Support non-Latin1 characters in CodeGenerator::visitFromCharCode()?

Status: NEW

(Core :: JavaScript Engine: JIT, enhancement, P3)

Description

[André Bargull [:anba]]

CodeGenerator::visitFromCharCode() currently calls into the VM when the input code unit is larger than 0xff (StaticStrings::UNIT_STATIC_LIMIT), which means |String.fromCharCode|, |String.prototype.charAt|, and indexed string accesses all call into the VM for non-Latin1 code units. CodeGenerator::visitFromCodePoint() supports non-Latin1 code points, we should measure if there are any negative side-effects to extend CodeGenerator::visitFromCharCode() to support non-Latin1 code units, too. 

|stringCodePoint| is 30-40% faster in this µ-benchmark when compared to |stringIndexed|, because String.fromCodePoint() supports non-Latin1 code points:
---
function stringIndexed() {
    var s = "abcdefg\u0100";
    var q = 0;
    var t = dateNow();
    for (var i = 0; i < 5000000; ++i) {
        for (var j = 0; j < s.length; ++j) {
            q += s[j].length;
        }
    }
    return [dateNow() - t, q];
}

function stringCodePoint() {
    var s = "abcdefg\u0100";
    var q = 0;
    var t = dateNow();
    for (var i = 0; i < 5000000; ++i) {
        for (var j = 0; j < s.length; ++j) {
            q += String.fromCodePoint(s.charCodeAt(j)).length;
        }
    }
    return [dateNow() - t, q];
}
---