Microsoft.IO.RecyclableMemoryStream VS fnlc

There’s also RecyclableMemoryStream

Another suggestion to try, there is a tool provided by Microsoft called Microsoft.IO.RecyclableMemoryStream which greatly reduces the amount of memory to garbage collect when streaming large amounts of data.

it's also common, however, that in GC'd languages people end up making their own mini-allocators to avoid producing garbage in the first place. See, for example, https://learn.microsoft.com/en-us/dotnet/api/system.buffers.arraypool-1?view=net-6.0, which didn't exist in Net Framework but Net Core added because even with a good GC, not GC'ing is faster than GC'ing. Or check out https://github.com/Microsoft/Microsoft.IO.RecyclableMemoryStream, which is basically recreating the pooling strategy that good Rust allocators also use (albeit with fewer pools because it's less general-purpose).

There are some known performance issues when using memorystreams. Microsoft themselves created a nuget package as a drop in replacement: https://github.com/microsoft/Microsoft.IO.RecyclableMemoryStream They talk about some of the issues in the readme, I suggest you take a look and see if anything applies to your use case.

If you do need to use MemoryStream a lot or reuse one, I suggest you to use https://github.com/Microsoft/Microsoft.IO.RecyclableMemoryStream it pools MemoryStream for you to optimize things

I had implemented logging of REST API requests to public service and one problem was pooling streams for reading request body. So I have found package Microsoft.IO.RecyclableMemoryStream. I'm using it following manner

.net 6 with AVX: https://github.com/hyrmn/fnlc