In a literal sense, the LLVM IR is more informative because it literally has more information attached to the instructions. This information gets lost when the

In a literal sense, the LLVM IR is more informative because it literally has more information attached to the instructions. This information gets lost when the machine code is produced. This information will be very useful to any compiler that uses LLVM as a back end, not just LLVM itself. Of course it may not be particularly useful to anyone else unless they're particularly interested in trying to diagnose why their code is being optimized in a certain way.