stream: writable buffering

[ronag]

Simplify and optimize writable stream buffering.

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• tests and/or benchmarks are included
• documentation is changed or added
• commit message follows commit guidelines

[ronag]

@mcollina @Trott I need a little help on how to run and compare benchmarks on this? Haven't really found the instructions in the guidelines...

[mcollina]

Here they are: https://github.com/nodejs/node/blob/master/doc/guides/writing-and-running-benchmarks.md

Be careful with this change, you are touching one of the worst hot paths in Node.js (together with emit  and nextTick). The current code is complex for reasons that made sense in old version of Node.js, so it might be a good time to check if those assumptions still hold true.

[mcollina]

The semversiness of this is a bit problematic, as this is likely to break some external users.

[ronag]

@mcollina: this is definitly semver major.

Is this something that you would even consider? If not please let me know so I don't spend time in vain :).

[mcollina]

I think it's worthwhile having a look if it could be simplified/improved the performance of it.

[lpinca]

This should definitely be benchmarked to see if it's actually faster and how much.

[ronag]

@mcollina @lpinca

Before:

writable-manywrites.js: 2,549,442

After:

writable-manywrites.js: 2,652,592

[ronag]

Got the benchmark comparison working!

streams/writable-manywrites.js n=2000000                 0.83 %       ±2.70% ±3.60% ±4.71%

Seems like a negligible difference. Though the simplification part is still probably worth it.

[ronag]

@mcollina nothing I do here seems to make much of a difference... maybe not a hot path or maybe better benchmarks are needed? Or I'm doing something wrong...

The current benchmark mostly test sync writing... which I don't think is a good real-world example.

[ronag]

Note this also changes the behaviour so that 'finish' is never emitted after 'error'.

[ronag]

@mcollina can you please look into this one #28979 before further reviewing this PR.

[ronag]

this is WIP