lib: improve priority queue performance

[falsandtru]

                                  confidence improvement accuracy (*)   (**)  (***)
util/priority-queue.js n=1000000        ***      6.66 %       ±2.93% ±3.90% ±5.08%

[falsandtru]

Path: parallel/test-vm-timeout-escape-promise-module
Error: --- stderr ---
(node:130552) ExperimentalWarning: VM Modules is an experimental feature. This feature could change at any time

This test often makes false positive #41839.

[falsandtru]

Can you approve?

[BridgeAR]

LGTM while the longer version suggested by @devsnek is still nicer to read for me and I would rather keep the benchmark runs, if possible.

[BridgeAR]

Do we have to increase the value? Our benchmarks have the tendency of taking very long overall.

[falsandtru]

The value is needed to get *** confidence stably. And the benchmark only takes one minute even after the change.

                                 confidence improvement accuracy (*)   (**)  (***)
util/priority-queue.js n=100000                 0.46 %       ±2.98% ±3.98% ±5.20%

                                 confidence improvement accuracy (*)   (**)  (***)
util/priority-queue.js n=100000        ***      4.36 %       ±1.99% ±2.65% ±3.46%

                                  confidence improvement accuracy (*)   (**)  (***)
util/priority-queue.js n=1000000        ***      9.43 %       ±3.13% ±4.17% ±5.43%

[falsandtru]

I know such a common static code can commonly be compact as an idiom. How about, other members?

[mscdex]

https://ci.nodejs.org/view/Node.js%20benchmark/job/benchmark-node-micro-benchmarks/1185/ and https://ci.nodejs.org/view/Node.js%20benchmark/job/benchmark-node-micro-benchmarks/1186/ don't seem to be able to reproduce the improvement.

[falsandtru]

Since those results have no confidence, the sample size is probably insufficient. Let's set runs to 100 as my previous pr did.

https://ci.nodejs.org/view/Node.js%20benchmark/job/benchmark-node-micro-benchmarks/1180/

Anyway the default sample size(runs value) 30 is too small. It have to be 60 or more.

[mscdex]

I re-ran it with 100 iterations and now it looks the same or worse:

https://ci.nodejs.org/view/Node.js%20benchmark/job/benchmark-node-micro-benchmarks/1188/console

util/priority-queue.js n=1000000         **     -2.08 %       ±1.29% ±1.70% ±2.19%

https://ci.nodejs.org/view/Node.js%20benchmark/job/benchmark-node-micro-benchmarks/1189/console

util/priority-queue.js n=1000000                -1.36 %       ±1.50% ±1.99% ±2.55%

[falsandtru]

I found that CI uses Ubuntu 16 and gcc/g++8, but I'm using Ubuntu 20.04 and gcc/g++ 9.4.0. CI may be inaccurate because of its legacy environment. Can anyone run benchmark on Ubuntu 20 or later and gcc/g++ 9.4.0?

[falsandtru]

How to

1. git checkout main
2. ./configure && make -j4 && mv out/Release/node ./node-main
3. Apply the changes.
4. ./configure && make -j4 && mv out/Release/node ./node-pr
5. node benchmark/compare.js --old ./node-main --new ./node-pr --filter "prio" util > compare-pr.csv && node-benchmark-compare compare-pr.csv

$ node benchmark/compare.js --old ./node-main --new ./node-pr --filter "prio" util > compare-pr.csv && node-benchmark-compare compare-pr.csv
[00:01:05|% 100| 1/1 files | 60/60 runs | 1/1 configs]: Done
                                  confidence improvement accuracy (*)   (**)  (***)
util/priority-queue.js n=1000000        ***      9.66 %       ±2.80% ±3.72% ±4.84%

Be aware that when doing many comparisons the risk of a false-positive result increases.
In this case, there are 1 comparisons, you can thus expect the following amount of false-positive results:
  0.05 false positives, when considering a   5% risk acceptance (*, **, ***),
  0.01 false positives, when considering a   1% risk acceptance (**, ***),
  0.00 false positives, when considering a 0.1% risk acceptance (***)

$ g++ --version
g++ (Ubuntu 9.4.0-1ubuntu1~20.04.1) 9.4.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

$ gcc --version
gcc (Ubuntu 9.4.0-1ubuntu1~20.04.1) 9.4.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

[mscdex]

I ran the benchmark several times locally for 100 iterations each and on my machine (with gcc 9.4.0) I consistently get results like:

util/priority-queue.js n=1000000        ***     -4.60 %       ±0.15% ±0.20% ±0.26%

At any rate, it seems the code change here isn't a clear performance win.

[falsandtru]

I couldn't reproduce the improvement on GitHub Actions. I've no idea what made the difference.