Hardware acceleration for CoreML

Hello
I'd like to aquire a full license of Parallels but I've got a question before.
We're developping an application which uses neural negine CoreML Apple framework for image recognition.
On a normal Mac, it uses neural engine to make inference. I've started our application in a parallels VM hosted on a Sillicon Mac and the inference works, but seem very slow, like if it was not using hardware but software.
Do you know if there is a way to make work it or if it is planned ?

Best regards
Cesar