[clang] [llvm] AMDGPU: Give v_cvt_sr_pk_bf16_f32 its own subtarget feature (PR #213150)

[Matt Arsenault via cfe-commits]

================
@@ -506,6 +506,7 @@ static void fillAMDGCNFeatureMap(StringRef GPU, const Triple &T,
     Features["tensor-cvt-lut-insts"] = true;
     Features["bf16-trans-insts"] = true;
     Features["bf16-cvt-insts"] = true;
+    Features["cvt-sr-pk-bf16-f32-inst"] = true;
----------------
arsenm wrote:

That's already tested 

https://github.com/llvm/llvm-project/pull/213150