Memory spec question

Reference

Q: This line:
...at the cost of significant performance penalties for reading and writing to the memory.

What is it about using Memory over Span that incurs "significant performance penalties"? This is not explained and I can't find anything anywhere about this.

[ahsonkhan]

If you want to read/write from/to a Memory, you have to first access its Span property and then index into it/etc. Accessing the property, though optimized recently (dotnet/coreclr#20386 / dotnet/coreclr#20771), isn't free (especially if you are going through the MemoryManager), and that's where part of the performance cost comes from. The Memory struct is also larger so passing it around/slicing it/etc. are generally slower too. Initially the delta was more noticeable, but it exists none the less.

https://github.com/dotnet/corefx/blob/908cba9d8f275935a9e91b32fd363b5687dc8f86/src/Common/src/CoreLib/System/Memory.cs#L293-L375

Certainly you wouldn't want to loop through the memory like this:

for (int i = 0; i < memory.Length; i++)
{
   sum += memory.Span[i];
}

But rather like this:

var span = memory.Span;
for (int i = 0; i < span.Length; i++)
{
   sum += span[i];
}

I can't find anything anywhere about this.

Maybe this article would help: https://www.codemag.com/article/1807051/Introducing-.NET-Core-2.1-Flagship-Types-Span-T-and-Memory-T

• See Let’s Take a Peek Inside Memory and Tricks of the Trade (particularly tip 1)

Great info thx👍