[llvm] [AMDGPU] Add subtarget features for MAD NC and 64-bit MIN/MAX instructions (PR #196326)

[Matt Arsenault via llvm-commits]

================
@@ -1317,10 +1317,18 @@ defm MadU32Inst : AMDGPUSubtargetFeature<"mad-u32-inst",
   "Has v_mad_u32 instruction"
 >;
 
+defm MadNC64_32Insts : AMDGPUSubtargetFeature<"mad-nc-64-32-insts",
+  "Has  v_mad_nc_{u64_u32|i64_i32} instructions"
+>;
+
 defm AddMinMaxInsts : AMDGPUSubtargetFeature<"add-min-max-insts",
   "Has v_add_{min|max}_{i|u}32 instructions"
 >;
 
+defm MinMax64Insts : AMDGPUSubtargetFeature<"min-max-64-insts",
----------------
arsenm wrote:

s/64/i64/ would help 

https://github.com/llvm/llvm-project/pull/196326