[llvm] [AMDGPU] Handle vector types for reqd_work_group_size constant folding. (PR #179551)

[Marcos Maronas via llvm-commits]

maarquitos14 wrote:

> I still do not think it makes any sense to make this conditional on being a vector type. This should be an expression of unconditionally trying to coerce it to the scalar integer type

If I understand correctly what you mean, you want the bitcast to be unconditional. However, if we do that, we run into assertions failing for vectors of more than one element (e.g. <2 x i8>). Every data type that has a total size of 16 bits can end up in this code, that's why we need the check for the bitcast, IMO. 

https://github.com/llvm/llvm-project/pull/179551