[llvm-dev] Canonicalize induction variables

[Sanjoy Das via llvm-dev]

Hi,

I have a "back to square one question -- why do you care about
having a canonical induction variable?

  - If it is for easier analysis, a somewhat better approach is to lean
    on SCEV to "understand" induction variables as much as possible.

  - If it is for better codegen, then this should be solved in
    LoopStrengthReduce.

That is not to say you're wrong in demanding this of LLVM, but I'd
like to understand your motivation a little better.

-- Sanjoy