[QUERY] [EventHub] PartitionKey HashCode Generation

Library name and version

Azure.Messaging.EventHubs

Query/Question

When converting PartitionResolver.ComputeHash to span and BinaryPrimitives it would be possible to squeeze out some perf. See the below benchmark

BenchmarkDotNet=v0.13.1, OS=Windows 10.0.22000
11th Gen Intel Core i7-1185G7 3.00GHz, 1 CPU, 8 logical and 4 physical cores
.NET SDK=6.0.102
  [Host]     : .NET 6.0.2 (6.0.222.6406), X64 RyuJIT
  DefaultJob : .NET 6.0.2 (6.0.222.6406), X64 RyuJIT

Method	Mean	Error	StdDev	StdErr	Median	Min	Q1	Q3	Max	Op/s	Ratio	RatioSD
ComputeHashBitConverter	9.599 ns	0.3469 ns	1.007 ns	0.1022 ns	9.421 ns	7.989 ns	8.835 ns	10.279 ns	12.61 ns	104,174,105.6	1.00	0.00
ComputeHashSpan	8.792 ns	0.4335 ns	1.216 ns	0.1274 ns	8.444 ns	6.987 ns	7.875 ns	9.409 ns	12.60 ns	113,736,404.7	0.93	0.16

Is that something you’d be interested in?

Another thing. I noticed currently it is required to allocate a byte array to compute the hash and the size of the byte array is proportional to the size of the partition key. I see the following options:

• Pool the buffer
• Stackalloc the buffer for short string and for longer pool the buffer
• If the EnqueueEventOptions are most likely to be reused per partition, it would be possible to turn the partition key into bytes there and the just pass those bytes into ComputeHash method (then the options act as sort of root for the byte array)

Just as an example, pooling the byte array there is slightly slower but gets rid of all the allocations on that path.

BenchmarkDotNet=v0.13.1, OS=Windows 10.0.22000
11th Gen Intel Core i7-1185G7 3.00GHz, 1 CPU, 8 logical and 4 physical cores
.NET SDK=6.0.102
  [Host]     : .NET 6.0.2 (6.0.222.6406), X64 RyuJIT
  DefaultJob : .NET 6.0.2 (6.0.222.6406), X64 RyuJIT

Method	Mean	Error	StdDev	StdErr	Min	Q1	Median	Q3	Max	Op/s	Ratio	RatioSD	Gen 0	Allocated
Pooled	67.84 ns	1.171 ns	1.038 ns	0.277 ns	66.13 ns	67.12 ns	67.67 ns	68.67 ns	69.39 ns	14,740,063.2	1.15	0.09	-	-
NonPooled	55.45 ns	1.686 ns	4.892 ns	0.497 ns	47.37 ns	51.71 ns	54.57 ns	58.57 ns	68.62 ns	18,033,322.7	1.00	0.00	0.0102	64 B

Environment

No response

[jsquire]

As always, thank you, @danielmarbach. I’m going to mark this one as addressed, but feel free to reach out if there’s any additional context that I can help with.

[danielmarbach]

Cool will think about it and once I have sufficient time I’ll send a PR