[llvm] [RFC] implement convergence control in MIR using SelectionDAG (PR #71785)

[Sameer Sahasrabuddhe via llvm-commits]

ssahasra wrote:

Bump! The only TODO is to mark every convergent function with HasPostISelHook = 1 in the TD files for AMDGPU. I am working on cleaner ways of doing this, but it need not block this review. Having the actual implementation of MIR tokens in upstream will be very useful.

https://github.com/llvm/llvm-project/pull/71785