System.Diagnostics.Activity enumeration benchmarks

[CodeBlanch]

Adding the benchmarks used on dotnet/runtime#40544

/cc @noahfalk @tarekgh

[tarekgh]

for [](start = 12, length = 3)

do we really need to loop? Benchmark already repeat the method execution anyway and we can get he exact number for the enumeration.

I would suggest too to create multiple static activities with different tags count. we can have the perf numbers for short, medium and long tags count.

[CodeBlanch]

Removed the looping. Added small/large tests.

[tarekgh]

OTelHelperActivityTagObjects [](start = 20, length = 28)

could you please use the full word OpenTelemetry instead?

[tarekgh]

Also, are we using OpenTelemetry because this how it call the Activity for enumeration? Does OTel create Enumerator as this test does? I am asking to know if it is worth it to complicate the test here. I am not pushing back though.

---

In reply to: 473440210 [](ancestors = 473440210)

[CodeBlanch]

Removed all the OpenTelemetry stuff.

[tarekgh]

Enumerator [](start = 22, length = 10)

I am not really sure if it is good to test the internals with the reflection and code emit. what is really the goal here?

[CodeBlanch]

These benchmarks were written for the other PR, really. I needed to make sure the struct enumerators worked for completely removing the allocation with the reflection mechanism, and also show there was an allocation reduction for the standard enumeration too. I opened this PR because I think @noahfalk said it would be useful. So my goal now is just to provide usefulness to future users of Activity 😄 I can remove the OpenTelemetry stuff, or I can close the PR completely if we don't think it will be that useful. LMK

[tarekgh]

Adding perf tests for Activity is useful. I suggest removing all OTel stuff and keep the simple test cases for the enumerating tags, links...etc. Thanks a lot for all your provided help.

[tarekgh]

CC @adamsitnik if he has any comment.

[tarekgh]

@CodeBlanch thanks for updating it. LGTM. I assume you ran it and the output is reasonable to you. right.

@adamsitnik @DrewScoggins looks I don't have a permission to merge in this repo. could you please have a look and merge it if you don't have any comment? Thanks.

[CodeBlanch]

@tarekgh Ya it's running fine. Here's the output:

Method	Mean	Error	StdDev	Median	Min	Max	Gen 0	Gen 1	Gen 2	Allocated
EnumerateActivityTagsSmall	64.20 ns	2.596 ns	2.990 ns	64.97 ns	59.36 ns	70.10 ns	0.0065	-	-	56 B
EnumerateActivityTagsLarge	10,649.76 ns	1,301.307 ns	1,498.587 ns	9,959.51 ns	9,232.34 ns	13,946.43 ns	-	-	-	56 B
EnumerateActivityTagObjectsSmall	89.92 ns	3.850 ns	4.434 ns	88.89 ns	83.85 ns	98.86 ns	0.0044	-	-	40 B
EnumerateActivityTagObjectsLarge	16,737.40 ns	808.777 ns	794.327 ns	16,513.94 ns	15,501.91 ns	18,173.94 ns	-	-	-	40 B
EnumerateActivityLinksSmall	116.74 ns	0.812 ns	0.634 ns	116.70 ns	115.73 ns	117.67 ns	0.0072	-	-	64 B
EnumerateActivityLinksLarge	22,298.54 ns	409.880 ns	383.402 ns	22,225.34 ns	21,698.86 ns	22,929.51 ns	-	-	-	64 B
EnumerateActivityEventsSmall	88.37 ns	1.154 ns	1.080 ns	88.34 ns	86.60 ns	90.12 ns	0.0065	-	-	56 B
EnumerateActivityEventsLarge	18,992.88 ns	290.204 ns	271.457 ns	19,048.33 ns	18,594.74 ns	19,382.68 ns	-	-	-	56 B
EnumerateActivityLinkTagsSmall	89.10 ns	9.694 ns	10.775 ns	85.42 ns	74.65 ns	114.33 ns	0.0057	-	-	48 B
EnumerateActivityLinkTagsLarge	14,855.37 ns	1,645.641 ns	1,760.818 ns	14,261.70 ns	13,465.29 ns	19,684.47 ns	-	-	-	48 B

[DrewScoggins]

LGTM