timers: use ref counts to count timers

[RaisinTen]

The additional objects that were getting added and deleted from the
activeTimersMap object were slowing down the rest of the timers code,
so this change falls back to using the ref counts to count the active
timers inside process.getActiveResourcesInfo().

Fixes: #41219

Signed-off-by: Darshan Sen darshan.sen@postman.com

cc @mscdex @nodejs/timers

[nodejs-github-bot]

Review requested:

• @nodejs/startup

[RaisinTen]

The benchmark results look very similar to the results of the revert PR - #41298. I think we should go ahead with this change because the slowdown in timers/timers-breadth-args.js and timers/timers-insert-pooled.js is present in the revert PR too.

[apapirovski]

I left some comments but honestly they're all superfluous because this and the original PR are both doing entirely too much work. Both Timer & Immediate already do their own book-keeping in JS-land for the number of active objects. You just need to export a function that returns refCount & immediateInfo[kRefCount] from internal/timers.js.

[RaisinTen]

@apapirovski thanks for the review! I've updated the PR to use refCount and immediateInfo[kRefCount] to count the timers, PTAL.

[RaisinTen]

Timers benchmark CI: https://ci.nodejs.org/job/benchmark-node-micro-benchmarks/1080/

 confidence improvement accuracy (*) (**) (***)
timers/immediate.js type='breadth1' n=5000000 -1.17 % ±4.24% ±5.65% ±7.38%
timers/immediate.js type='breadth4' n=5000000 ** 6.94 % ±4.99% ±6.67% ±8.76%
timers/immediate.js type='breadth' n=5000000 *** 19.86 % ±5.15% ±6.88% ±8.98%
timers/immediate.js type='clear' n=5000000 *** 282.93 % ±4.54% ±6.11% ±8.11%
timers/immediate.js type='depth1' n=5000000 *** 41.79 % ±1.93% ±2.58% ±3.38%
timers/immediate.js type='depth' n=5000000 *** 41.20 % ±1.26% ±1.67% ±2.17%
timers/set-immediate-breadth-args.js n=5000000 -0.07 % ±4.01% ±5.35% ±6.99%
timers/set-immediate-breadth.js n=10000000 -1.82 % ±3.54% ±4.71% ±6.15%
timers/set-immediate-depth-args.js n=5000000 *** 39.03 % ±2.49% ±3.31% ±4.31%
timers/timers-breadth-args.js n=1000000 *** -12.75 % ±5.98% ±8.00% ±10.51%
timers/timers-breadth.js n=5000000 2.47 % ±3.74% ±4.98% ±6.50%
timers/timers-cancel-pooled.js n=5000000 *** 188.38 % ±24.54% ±32.95% ±43.50%
timers/timers-cancel-unpooled.js direction='end' n=1000000 ** 24.24 % ±15.35% ±20.48% ±26.75%
timers/timers-cancel-unpooled.js direction='start' n=1000000 * 9.86 % ±8.87% ±11.81% ±15.39%
timers/timers-depth.js n=1000 -0.24 % ±0.47% ±0.63% ±0.82%
timers/timers-insert-pooled.js n=5000000 *** -17.76 % ±3.28% ±4.36% ±5.68%
timers/timers-insert-unpooled.js direction='end' n=1000000 *** 11.50 % ±4.24% ±5.65% ±7.36%
timers/timers-insert-unpooled.js direction='start' n=1000000 *** 10.90 % ±3.99% ±5.31% ±6.91%
timers/timers-timeout-nexttick.js n=50000 *** 18.12 % ±6.26% ±8.35% ±10.93%
timers/timers-timeout-nexttick.js n=5000000 *** 70.62 % ±8.74% ±11.63% ±15.16%
timers/timers-timeout-pooled.js n=10000000 *** 93.09 % ±11.97% ±16.06% ±21.16%
timers/timers-timeout-unpooled.js n=1000000 *** 88.09 % ±14.55% ±19.37% ±25.22%

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/41707/

[apapirovski]

Looks good. Thanks for making the change!

[aduh95]

Benchmark results look impressive 👏

[nodejs-github-bot]

Landed in 64cc066