Remove use of Tuple<> from corelib

[stephentoub]

Related to #44684
cc: @jkotas, @marek-safar

(The flip side of this change, of course, is that if Tuple<T1, T2> is already being used by the rest of the app, this could actually increase the overall size rather than decrease it.)

[Dotnet-GitSync-Bot]

I couldn't figure out the best area label to add to this PR. If you have write-permissions please help me learn by adding exactly one area label.

[jkotas]

I think this is a good idea. If we take this route, we should also do a pass to remove uses of Tuple in netcoreapp that are easy to avoid. For example, the Tuple use in src\libraries\System.Security.Cryptography.Algorithms\src\Internal\Cryptography\AesImplementation.Unix.cs can be replaced with ValueTuple or custom type.

[stephentoub]

we should also do a pass to remove uses of Tuple in netcoreapp that are easy to avoid

For uses that need to pass a few arguments through an object state, is there a measurable downside to boxing/unboxing a ValueTuple<> instead of using Tuple<>?

[jkotas]

is there a measurable downside to boxing/unboxing

I doubt it that the difference is measurable. It is just an extra local copy (ie a few extra mov instructions).

[jkotas]

There is a difference between TupleSlim vs. Tuple/ValueTuple for AOT. The full AOT compilers will typically end up generating the code for all interfaces and methods on Tuple/ValueTuple. TupleSlim avoids this overhead.

[stephentoub]

There is a difference between TupleSlim vs. Tuple/ValueTuple for AOT

Right.

for all interfaces and methods on Tuple/ValueTuple

Minus anything that could be trimmed, but presumably the problem here is little can be trimmed because most of the methods are interface implementations or virtual overrides.

It is just an extra local copy

And if it were an issue (which it shouldn't be), there's always Unsafe.Unbox I guess.