Investigate some potential perf improvements in Number.BigInteger.cs

[tannergooding]

The floating-point formatting/parsing code currently uses an internal BigInteger ref struct that would likely benefit from using specific hardware intrinsics.

We should investigate using the BMI intrinsics, where possible and also look at exposing new intrinsics around AddWithCarry and Multiply128, both of which can simplify some of the more complex codegen down to a single instruction. You can see dotnet/coreclr#22040 (comment) for some more details and the surrounding discussion.

[mazong1123]

This sounds an interesting work item to me. If no one is working on it I'd like to take a look.

I went through the code for HW intrinsics, seems to add a new intrinsic we need to touch System/Runtime/Intrinsics and hwintrinsiclistxarch.h. Not sure if there's any guidance. If no I can figure it out as well.

[mazong1123]

Looking at the board of HW intrinsics, seems an API proposal is needed before implementation. If we decide to add new intrinsics, I would create separate issues for proposing them first. Not sure if this is the right process.

[tannergooding]

Since the BigInteger ref struct exists within S.P.Corelib, the AddWithCarry and Multiply128 intrinsics could be prototyped without going through an API proposal. That is, they can be kept internal and experimented with to get numbers and prove they are useful first.

Tagging subscribers to this area: @tannergooding
Notify danmosemsft if you want to be subscribed.

[dotnet-policy-service]

Due to lack of recent activity, this issue has been marked as a candidate for backlog cleanup. It will be closed if no further activity occurs within 14 more days. Any new comment (by anyone, not necessarily the author) will undo this process.

This process is part of our issue cleanup automation.