Should we declare the instructions set used by some QIR code?

[LaurentAjdnik]

As part of a "target duo" (similar to LLVM's "target triple"), it is suggested that the instructions set for which some code has been compiled into QIR should be mentioned inside the LLVM file.

Would it be a comprehensive list of all targeted functions?

Or a set identifier (which would then need to be inventoried somewhere else)?

Does it make sense to mention a set of instructions if only a few ones might actually be used along the code?

In fact, is it even necessary, since all the required QIS functions will be declared as external functions in the LLVM code and can easily be rebuilt on the flight by an "analysis pass"?