perf(ext/websocket): use opAsync2 to avoid spread deopt

[littledivy]

This commit adds a new core API opAsync2 to call an async op with atmost 2 arguments. Spread argument iterators has a pretty big perf hit when calling ops.

name	avg msg/sec/core
1.32.1	127820.750000
#18506	140079.000000
#18506 + #18509	150104.250000
#18506 + #18509 + this	157340.000000

[aapoalas]

Looks good to me. One thing I wonder is what was the performance impact of doing async ops with a callback API? Ie. Create a CB taking the promiseid parameter that core then synchronously calls to init the actual op.

[petamoriken]

Maybe switch (args.length) {} is good for performance in opAsync.

[bartlomieju]

Please migrate all relevant ops to this variant

[littledivy]

@aapoalas For promises, we currently batch multiple completed operations in one single opresolve call. With callbacks for each one, it will be multiple entries into JS. I think we did investigate this some time ago.

[aapoalas]

@aapoalas For promises, we currently batch multiple completed operations in one single opresolve call. With callbacks for each one, it will be multiple entries into JS. I think we did investigate this some time ago.

No I mean for starting the op: What the opAsync call essentially does is that it injects the promiseid argument to the actual call's arguments. If instead it took as parameter a single callback that it called with the promiseid then the caller of opAsync could in that callback call the actual op binding function without spreading args.

[littledivy]

No I mean for starting the op: What the opAsync call essentially does is that it injects the promiseid argument to the actual call's arguments. If instead it took as parameter a single callback that it called with the promiseid then the caller of opAsync could in that callback call the actual op binding function without spreading args.

Yup that would work. We could also try arguments.length if thats faster. For now, I like opAsync2, opAsync3 because its easier to mass replace.

[bartlomieju]

LGTM