- // Initialize the parser with a fresh subtarget info. It is better to use a
- // new STI here because the parser may modify it and we do not want those
- // modifications to persist after parsing the inlineasm. The modifications
- // made by the parser will be seen by the code emitters because it passes
- // the current STI down to the EncodeInstruction() method.
- std::unique_ptr<MCSubtargetInfo> STI(TM.getTarget().createMCSubtargetInfo(
- TM.getTargetTriple(), TM.getTargetCPU(), TM.getTargetFeatureString()));
-
- // Preserve a copy of the original STI because the parser may modify it. For
- // example, when switching between arm and thumb mode. If the target needs to
- // emit code to return to the original state it can do so in
+ // Create a temporary copy of the original STI because the parser may modify
+ // it. For example, when switching between arm and thumb mode. If the target
+ // needs to emit code to return to the original state it can do so in