stream: some simplification

[ronag]

~7% improvement

NOTE: Contains 2 commits to reduce the number of PRs.

 streams/writable-manywrites.js callback='no' writev='no' sync='no' n=2000000                    6.27 %       ±6.47%  ±8.63% ±11.26%
 streams/writable-manywrites.js callback='no' writev='no' sync='yes' n=2000000            *      8.43 %       ±6.68%  ±8.90% ±11.59%
 streams/writable-manywrites.js callback='no' writev='yes' sync='no' n=2000000                   7.77 %       ±7.88% ±10.49% ±13.69%
 streams/writable-manywrites.js callback='no' writev='yes' sync='yes' n=2000000           *      6.79 %       ±5.75%  ±7.65%  ±9.96%
 streams/writable-manywrites.js callback='yes' writev='no' sync='no' n=2000000          ***     11.31 %       ±6.08%  ±8.11% ±10.60%
 streams/writable-manywrites.js callback='yes' writev='no' sync='yes' n=2000000          **      7.76 %       ±5.46%  ±7.27%  ±9.46%
 streams/writable-manywrites.js callback='yes' writev='yes' sync='no' n=2000000                  6.12 %       ±7.40%  ±9.85% ±12.82%
 streams/writable-manywrites.js callback='yes' writev='yes' sync='yes' n=2000000         **      8.11 %       ±4.90%  ±6.52%  ±8.48%

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

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/30806/

[ronag]

This condition is a duplicate of what's inside of clearBuffer. However, it is here on purpose since clearBuffer is not inlineable.

[mscdex]

This seems to have a negative effect on Duplex creation:

streams/creation.js kind='duplex' n=50000000                                           ***     -6.94 %       ±3.44% ±4.62%  ±6.09%

[ronag]

@mscdex Flaky made?

Local:

 streams/creation.js kind='duplex' n=5000000                 2.88 %       ±5.44% ±7.25% ±9.46%

[ronag]

Benchmark CI (creation): https://ci.nodejs.org/view/Node.js%20benchmark/job/benchmark-node-micro-benchmarks/556/

[ronag]

Nope, seems consistent:

17:55:08  streams/creation.js kind='duplex' n=50000000           ***     -7.40 %       ±2.49% ±3.31% ±4.31%

Strange... not sure how this PR could cause that

[ronag]

https://ci.nodejs.org/view/Node.js%20benchmark/job/benchmark-node-micro-benchmarks/557/

[ronag]

@mscdex: I'm quite baffled why this PR would affect duplex creation performance. Any ideas?

Anyway, a 7% performance increase in writing I guess is still worth it with a 7% performance degradation in creation?

[mscdex]

I don't know offhand, you'd probably have to compare either the generated/optimized code or optimizer/inliner trace output before and after this PR to see what is different.