[llvm] [AMDGPU][True16][MC] 16bit vsrc and vdst support in MC (PR #104510)

[Joe Nash via llvm-commits]

================
@@ -1165,8 +1165,8 @@ multiclass f16_fp_Pats<Instruction cvt_f16_f32_inst_e64, Instruction cvt_f32_f16
 let SubtargetPredicate = NotHasTrue16BitInsts in
 defm : f16_fp_Pats<V_CVT_F16_F32_e64, V_CVT_F32_F16_e64>;
 
-let SubtargetPredicate = HasTrue16BitInsts in
-defm : f16_fp_Pats<V_CVT_F16_F32_t16_e64, V_CVT_F32_F16_t16_e64>;
+let SubtargetPredicate = UseFakeTrue16Insts in
----------------
Sisyph wrote:

Use True16Predicate instead of SubtargetPredicate

https://github.com/llvm/llvm-project/pull/104510