[clang] [llvm] AMDGPU: Generate TargetParser table from TableGen (PR #212353)

[Alexis Engelke via llvm-commits]

aengelke wrote:

Note that this increases the critical path length of building LLVM with AMDGPU enabled, as almost everything (including IR) depends on TargetParser and AMDGPU.td takes *ages* to parse (12-13s on my system). Why does this need to use the entire AMDGPU.td? Can't the tablegen files be made more modular?

https://github.com/llvm/llvm-project/pull/212353