What is the correct method to model caches (data cache, instruction cache) while developing an LLVM backend. Is it done in one of the TableGen (.td
) files? How do Intel, RISC-V etc. do it?
As of writing this comment, there does not seem to be a standard method, but, the RFC [System Model] Introduce a target system model did propose a way. It seems to be still under review!
This post is related to my other question TableGen: Apart from registers, do we also model internal memory?