Mqtt311Decoder: use ImmutableList.Builder

[Aaronontheweb]

trafficstars

Fewer object allocations during packet parsing this way

[Aaronontheweb]

dev

BenchmarkDotNet v0.14.0, Windows 11 (10.0.22631.4037/23H2/2023Update/SunValley3)
12th Gen Intel Core i7-1260P, 1 CPU, 16 logical and 12 physical cores
.NET SDK 8.0.302
  [Host]     : .NET 8.0.6 (8.0.624.26715), X64 RyuJIT AVX2
  Job-GKHTFW : .NET 8.0.6 (8.0.624.26715), X64 RyuJIT AVX2

InvocationCount=1  LaunchCount=10  RunStrategy=Monitoring  
UnrollFactor=1  WarmupCount=10  

Method	QoSLevel	PayloadSizeBytes	ProtocolVersion	Mean	Error	StdDev	Median	Req/sec
PublishAndReceiveMessages	AtMostOnce	10	V3_1_1	3.519 μs	0.7376 μs	2.175 μs	2.758 μs	284,182.50
PublishAndReceiveMessages	AtMostOnce	1024	V3_1_1	3.811 μs	0.6556 μs	1.933 μs	3.095 μs	262,409.13
PublishAndReceiveMessages	AtMostOnce	2048	V3_1_1	4.430 μs	0.6173 μs	1.820 μs	3.831 μs	225,738.78
PublishAndReceiveMessages	AtMostOnce	8192	V3_1_1	13.250 μs	2.4129 μs	7.114 μs	9.620 μs	75,470.25
PublishAndReceiveMessages	AtLeastOnce	10	V3_1_1	23.111 μs	1.9651 μs	5.794 μs	24.182 μs	43,268.72
PublishAndReceiveMessages	AtLeastOnce	1024	V3_1_1	23.023 μs	1.9696 μs	5.807 μs	24.033 μs	43,434.29
PublishAndReceiveMessages	AtLeastOnce	2048	V3_1_1	27.093 μs	2.9436 μs	8.679 μs	28.678 μs	36,910.36
PublishAndReceiveMessages	AtLeastOnce	8192	V3_1_1	32.266 μs	2.3566 μs	6.948 μs	32.402 μs	30,991.97
PublishAndReceiveMessages	ExactlyOnce	10	V3_1_1	42.754 μs	2.3814 μs	7.022 μs	40.309 μs	23,389.38
PublishAndReceiveMessages	ExactlyOnce	1024	V3_1_1	41.514 μs	1.9078 μs	5.625 μs	39.755 μs	24,088.12
PublishAndReceiveMessages	ExactlyOnce	2048	V3_1_1	40.881 μs	5.7051 μs	16.822 μs	40.495 μs	24,461.15
PublishAndReceiveMessages	ExactlyOnce	8192	V3_1_1	52.581 μs	2.4027 μs	7.084 μs	51.804 μs	19,018.14

This PR

BenchmarkDotNet v0.14.0, Windows 11 (10.0.22631.4037/23H2/2023Update/SunValley3)
12th Gen Intel Core i7-1260P, 1 CPU, 16 logical and 12 physical cores
.NET SDK 8.0.302
  [Host]     : .NET 8.0.6 (8.0.624.26715), X64 RyuJIT AVX2
  Job-XKOQDD : .NET 8.0.6 (8.0.624.26715), X64 RyuJIT AVX2

InvocationCount=1  LaunchCount=10  RunStrategy=Monitoring  
UnrollFactor=1  WarmupCount=10  

Method	QoSLevel	PayloadSizeBytes	ProtocolVersion	Mean	Error	StdDev	Median	Req/sec
PublishAndReceiveMessages	AtMostOnce	10	V3_1_1	3.776 μs	0.7134 μs	2.103 μs	2.993 μs	264,832.54
PublishAndReceiveMessages	AtMostOnce	1024	V3_1_1	3.528 μs	0.6014 μs	1.773 μs	2.803 μs	283,415.14
PublishAndReceiveMessages	AtMostOnce	2048	V3_1_1	4.319 μs	0.6248 μs	1.842 μs	3.870 μs	231,522.29
PublishAndReceiveMessages	AtMostOnce	8192	V3_1_1	14.125 μs	2.9633 μs	8.737 μs	10.495 μs	70,798.58
PublishAndReceiveMessages	AtLeastOnce	10	V3_1_1	25.145 μs	1.2204 μs	3.598 μs	24.535 μs	39,768.96
PublishAndReceiveMessages	AtLeastOnce	1024	V3_1_1	22.786 μs	1.8937 μs	5.584 μs	24.131 μs	43,885.74
PublishAndReceiveMessages	AtLeastOnce	2048	V3_1_1	27.929 μs	1.5958 μs	4.705 μs	28.190 μs	35,805.37
PublishAndReceiveMessages	AtLeastOnce	8192	V3_1_1	28.807 μs	2.5123 μs	7.408 μs	28.615 μs	34,713.84
PublishAndReceiveMessages	ExactlyOnce	10	V3_1_1	39.247 μs	2.3468 μs	6.920 μs	39.792 μs	25,479.42
PublishAndReceiveMessages	ExactlyOnce	1024	V3_1_1	39.877 μs	2.1262 μs	6.269 μs	38.985 μs	25,077.39
PublishAndReceiveMessages	ExactlyOnce	2048	V3_1_1	43.217 μs	3.9505 μs	11.648 μs	42.917 μs	23,139.00
PublishAndReceiveMessages	ExactlyOnce	8192	V3_1_1	51.701 μs	2.3448 μs	6.914 μs	49.587 μs	19,341.96

Looks like about even to me, which isn't a surprise. We should probably add a benchmark that looks at the overhead of decoding multiple packets - that'd be a "cleaner" way of capturing this information.