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

[Matt Arsenault via cfe-commits]

================
@@ -20,12 +20,38 @@ class AMDGPUArchFeature<string spelling> {
 // R600-only.
 def R600_FEATURE_FMA : AMDGPUArchFeature<"R600_FEATURE_FMA">;
 
-// Marks a Processor/ProcessorModel record as a canonical TargetParser GPU.
-class AMDGPUGPUInfo {
+// AMDGCN.
----------------
arsenm wrote:

Not sure about this one, the point was to differentiate from r600 (though the prevailing trend is to call everything else amdgpu)

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