Maybe this has has already been covered, by I would not target LLVM IR in 2024.<p>I'd target MLIR (like Mojo does).<p>1. It's a much easier/better target to work with.<p>2. It's a strict super-set of LLVM IR.<p>3. Much better optimizations are possible that are specific to your language.<p>Separately, I'd love to have a Clojure-friendly interface to MLIR—whether via Jank or something else.
Is LLVM IR a stable target these days? I once heard of a project that got bit pretty hard with LLVM IR interface changes in the (not all that recent) past.