Thank you for this reply.
However, I've already read this thesis. (Mr. Doug Simon in Sun microsystems I mentioned was her student.)
This is a quite good article for reverse engineering but this is only useful for my alternative way which is trying to derive HL code from a machine assembly.
This thesis is a little bit old before SSA form becomes quite popular so does not deal with any PHI function.
For this reason, I can't get any useful info for constructing loops from LLVM IR.
Anyway, thank you for your concerns.
Seung J. Lee