[Question] How to use value produced by LLVM in higher level dialect?

Hi, when migrating our llvm-based ML infra into MLIR, we want to use some high level dialect to represent our computation. However, there exists many external invocations and our main computation process may depends on these external invocation results. Though I can represent them just with LLVM dialect, I do not know how to represent dependencies to these invocation outputs in high level dialects.