Add BatchSemaphore::close_no_scheduling_point

[sarsko]

Adds BatchSemaphore::close_no_scheduling_point.

The reason for adding this is that primitives which utilize multiple semaphores (in my case: mpsc) need to be able to close more than one semaphore without having more than one scheduling point. This is both for efficiency, and for not leaving a system which is partially closed.

The latter could have been solved with an atomic field as the source of truth on whether the system is closed or not, but I prefer exposing this method as this approach is more efficient than having needless scheduling points.

There is however one con of this, which is that after #212 gets merged, a switch becomes both a scheduling point and an "indicator of interest". If this PR gets merged, then any users of BatchSemaphore::close_no_scheduling_point will not indicate any interest on those BatchSemaphores. There's a couple potential solutions to this that come to mind:

1. Ignore it
2. Add a new Event which takes multiple Events. The code would then never use close, and instead do a single switch before calling close_no_scheduling_point on whichever BatchSemaphores it wants to close.
   • Rumor has it this would also be useful to solve the issue I point out in my comment in this comment: Next Event on Task struct #212 (comment)

There might be a case to expose multiple _no_scheduling_point functions. My stance is that we'll add those once we need them. If it becomes the case that we want multiple _no_scheduling_point functions, then I will probably suggest to refactor the code such that switch(Event) returns a Switched(Event) which would be passed by ownership to the _no_scheduling_point functions in order to make sure that a scheduling point has occurred and that a corresponding Event has been created. Again, that bridge is to be burned if that river is ever to be crossed.

---

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

[github-actions]

Benchmark for 0855531

Click to view benchmark

Test	Base	PR	%
buffer/pct	75.8±0.49ms	75.1±1.01ms	-0.92%
buffer/random	104.6±0.27ms	105.2±1.84ms	+0.57%
counter async/pct-narrow	1110.7±28.48µs	1117.1±31.74µs	+0.58%
counter async/pct-wide	11.0±0.09ms	11.0±0.08ms	0.00%
counter async/random-narrow	1404.5±17.02µs	1387.1±17.97µs	-1.24%
counter async/random-wide	4.3±0.06ms	4.3±0.06ms	0.00%
counter scaling async/RW/{tasks:1024,events:100000}	77.6±0.72ms	76.7±0.47ms	-1.16%
counter scaling async/RW/{tasks:128,events:100000}	37.6±0.11ms	37.5±0.28ms	-0.27%
counter scaling async/RW/{tasks:128,events:10000}	4.8±0.05ms	4.8±0.05ms	0.00%
counter scaling async/RW/{tasks:16,events:100000}	17.8±0.05ms	17.5±0.03ms	-1.69%
counter scaling async/RW/{tasks:16,events:10000}	2.0±0.01ms	1986.3±19.80µs	-0.69%
counter scaling async/RW/{tasks:16,events:1000}	441.2±9.48µs	425.2±5.75µs	-3.63%
counter scaling async/RW/{tasks:32,events:100000}	21.0±0.09ms	20.8±0.12ms	-0.95%
counter scaling async/RW/{tasks:32,events:10000}	2.5±0.02ms	2.5±0.01ms	0.00%
counter scaling async/RW/{tasks:32,events:1000}	711.2±12.27µs	696.3±10.40µs	-2.10%
counter scaling async/RW/{tasks:4,events:100000}	13.1±0.03ms	12.8±0.03ms	-2.29%
counter scaling async/RW/{tasks:4,events:10000}	1384.2±3.08µs	1361.7±3.81µs	-1.63%
counter scaling async/RW/{tasks:4,events:1000}	201.5±2.60µs	204.5±2.42µs	+1.49%
counter scaling async/RW/{tasks:64,events:100000}	27.3±0.74ms	27.0±0.04ms	-1.10%
counter scaling async/RW/{tasks:64,events:10000}	3.5±0.02ms	3.5±0.02ms	0.00%
counter scaling async/RW/{tasks:64,events:1000}	1219.8±30.99µs	1235.5±23.16µs	+1.29%
counter sync/pct-narrow	1106.3±18.19µs	1106.9±22.92µs	+0.05%
counter sync/pct-wide	11.0±0.07ms	11.0±0.07ms	0.00%
counter sync/random-narrow	1413.8±9.44µs	1392.9±8.00µs	-1.48%
counter sync/random-wide	4.3±0.07ms	4.4±0.06ms	+2.33%
create async/pct-narrow	880.9±8.13µs	873.1±9.20µs	-0.89%
create async/pct-wide	29.0±0.17ms	28.8±0.12ms	-0.69%
create async/random-narrow	847.1±12.41µs	843.9±12.72µs	-0.38%
create async/random-wide	16.9±0.22ms	16.8±0.09ms	-0.59%
create sync/pct-narrow	918.7±9.55µs	915.1±10.20µs	-0.39%
create sync/pct-wide	30.0±0.12ms	29.5±0.10ms	-1.67%
create sync/random-narrow	860.3±9.54µs	858.0±7.83µs	-0.27%
create sync/random-wide	18.7±0.44ms	18.1±0.07ms	-3.21%
lock scaling sync/RW/{tasks:1024,events:100000}	15.8±0.06s	15.7±0.07s	-0.63%
lock scaling sync/RW/{tasks:128,events:100000}	2.1±0.01s	2.1±0.01s	0.00%
lock scaling sync/RW/{tasks:128,events:10000}	199.2±0.16ms	198.1±0.61ms	-0.55%
lock scaling sync/RW/{tasks:16,events:100000}	293.5±0.24ms	294.1±1.59ms	+0.20%
lock scaling sync/RW/{tasks:16,events:10000}	29.3±0.05ms	29.4±0.08ms	+0.34%
lock scaling sync/RW/{tasks:16,events:1000}	3.0±0.10ms	3.0±0.01ms	0.00%
lock scaling sync/RW/{tasks:32,events:100000}	553.4±9.87ms	549.1±2.44ms	-0.78%
lock scaling sync/RW/{tasks:32,events:10000}	54.4±0.10ms	54.2±0.07ms	-0.37%
lock scaling sync/RW/{tasks:32,events:1000}	5.5±0.02ms	5.5±0.02ms	0.00%
lock scaling sync/RW/{tasks:4,events:100000}	88.8±0.34ms	90.0±0.14ms	+1.35%
lock scaling sync/RW/{tasks:4,events:10000}	8.9±0.02ms	9.0±0.02ms	+1.12%
lock scaling sync/RW/{tasks:4,events:1000}	952.7±2.83µs	963.4±3.72µs	+1.12%
lock scaling sync/RW/{tasks:64,events:100000}	1059.8±1.48ms	1051.9±2.27ms	-0.75%
lock scaling sync/RW/{tasks:64,events:10000}	103.8±0.73ms	102.7±0.11ms	-1.06%
lock scaling sync/RW/{tasks:64,events:1000}	9.7±0.03ms	9.6±0.04ms	-1.03%
lock sync/pct-narrow	3.1±0.19ms	3.1±0.02ms	0.00%
lock sync/pct-wide	21.9±0.56ms	22.2±0.18ms	+1.37%
lock sync/random-narrow	10.8±0.05ms	11.0±0.04ms	+1.85%
lock sync/random-wide	158.6±0.33ms	157.0±0.75ms	-1.01%

[github-actions]

Benchmark for 1c8dfda

Click to view benchmark

Test	Base	PR	%
buffer/pct	75.4±1.51ms	76.0±0.85ms	+0.80%
buffer/random	104.4±1.05ms	104.8±1.13ms	+0.38%
counter async/pct-narrow	1108.0±25.13µs	1110.2±27.17µs	+0.20%
counter async/pct-wide	11.0±0.19ms	10.9±0.16ms	-0.91%
counter async/random-narrow	1370.9±39.90µs	1430.2±24.14µs	+4.33%
counter async/random-wide	4.2±0.07ms	4.3±0.07ms	+2.38%
counter scaling async/RW/{tasks:1024,events:100000}	76.8±0.72ms	76.1±0.87ms	-0.91%
counter scaling async/RW/{tasks:128,events:100000}	36.0±0.35ms	36.3±0.20ms	+0.83%
counter scaling async/RW/{tasks:128,events:10000}	4.7±0.05ms	4.7±0.05ms	0.00%
counter scaling async/RW/{tasks:16,events:100000}	17.1±0.15ms	17.9±0.06ms	+4.68%
counter scaling async/RW/{tasks:16,events:10000}	1944.7±19.25µs	2.0±0.03ms	+2.84%
counter scaling async/RW/{tasks:16,events:1000}	432.1±7.42µs	435.9±6.29µs	+0.88%
counter scaling async/RW/{tasks:32,events:100000}	20.8±0.39ms	21.2±0.30ms	+1.92%
counter scaling async/RW/{tasks:32,events:10000}	2.5±0.02ms	2.5±0.02ms	0.00%
counter scaling async/RW/{tasks:32,events:1000}	700.8±13.05µs	692.2±12.88µs	-1.23%
counter scaling async/RW/{tasks:4,events:100000}	12.8±0.02ms	13.3±0.05ms	+3.91%
counter scaling async/RW/{tasks:4,events:10000}	1350.8±6.81µs	1396.1±15.66µs	+3.35%
counter scaling async/RW/{tasks:4,events:1000}	198.8±2.19µs	200.0±2.43µs	+0.60%
counter scaling async/RW/{tasks:64,events:100000}	26.5±0.26ms	27.0±0.11ms	+1.89%
counter scaling async/RW/{tasks:64,events:10000}	3.4±0.04ms	3.5±0.04ms	+2.94%
counter scaling async/RW/{tasks:64,events:1000}	1211.1±21.97µs	1207.4±23.07µs	-0.31%
counter sync/pct-narrow	1106.4±17.48µs	1099.7±22.02µs	-0.61%
counter sync/pct-wide	11.0±0.16ms	11.0±0.08ms	0.00%
counter sync/random-narrow	1383.9±12.45µs	1441.5±24.70µs	+4.16%
counter sync/random-wide	4.2±0.09ms	4.3±0.06ms	+2.38%
create async/pct-narrow	885.4±12.53µs	862.3±13.34µs	-2.61%
create async/pct-wide	28.7±0.47ms	28.4±0.42ms	-1.05%
create async/random-narrow	839.4±8.59µs	828.3±10.69µs	-1.32%
create async/random-wide	16.7±0.14ms	16.7±0.15ms	0.00%
create sync/pct-narrow	921.5±6.96µs	894.4±19.45µs	-2.94%
create sync/pct-wide	29.8±0.35ms	29.5±0.27ms	-1.01%
create sync/random-narrow	849.6±9.74µs	851.3±10.96µs	+0.20%
create sync/random-wide	17.8±0.23ms	18.1±0.07ms	+1.69%
lock scaling sync/RW/{tasks:1024,events:100000}	16.0±0.06s	16.2±0.06s	+1.25%
lock scaling sync/RW/{tasks:128,events:100000}	2.1±0.01s	2.1±0.01s	0.00%
lock scaling sync/RW/{tasks:128,events:10000}	199.6±0.75ms	198.3±2.07ms	-0.65%
lock scaling sync/RW/{tasks:16,events:100000}	292.5±2.08ms	293.9±0.43ms	+0.48%
lock scaling sync/RW/{tasks:16,events:10000}	29.3±0.15ms	29.3±0.03ms	0.00%
lock scaling sync/RW/{tasks:16,events:1000}	3.0±0.01ms	3.0±0.01ms	0.00%
lock scaling sync/RW/{tasks:32,events:100000}	549.4±5.17ms	551.5±1.16ms	+0.38%
lock scaling sync/RW/{tasks:32,events:10000}	54.4±0.26ms	54.4±0.08ms	0.00%
lock scaling sync/RW/{tasks:32,events:1000}	5.5±0.07ms	5.5±0.01ms	0.00%
lock scaling sync/RW/{tasks:4,events:100000}	89.1±0.33ms	88.1±0.07ms	-1.12%
lock scaling sync/RW/{tasks:4,events:10000}	8.9±0.06ms	8.8±0.02ms	-1.12%
lock scaling sync/RW/{tasks:4,events:1000}	956.2±7.22µs	948.1±8.13µs	-0.85%
lock scaling sync/RW/{tasks:64,events:100000}	1057.3±5.53ms	1058.4±5.78ms	+0.10%
lock scaling sync/RW/{tasks:64,events:10000}	103.5±0.88ms	103.7±0.07ms	+0.19%
lock scaling sync/RW/{tasks:64,events:1000}	9.7±0.08ms	9.7±0.03ms	0.00%
lock sync/pct-narrow	3.0±0.04ms	3.1±0.02ms	+3.33%
lock sync/pct-wide	21.7±0.16ms	21.4±0.14ms	-1.38%
lock sync/random-narrow	10.9±0.04ms	10.8±0.04ms	-0.92%
lock sync/random-wide	158.1±0.95ms	158.1±0.49ms	0.00%