Fwd: [RFC] How to manifest information in LLVM-IR, or, revisiting llvm.assume
I just posted an RFC to the llvm-dev list that might be interesting for
Clang and Flang folks as well. To keep the conversation on a single list
I only attached the abstract below. Please respond on llvm-dev, thanks!
> It is often hard or impossible to encode complex, e.g., non-boolean,
> information in an `llvm.assume(i1)`. This RFC describes various problems
> we have right now and provides alternative design ideas.