What you did with McSema was really impressive and I’m glad discussing with one of the authors! I get the point you regret using LLVM, it seemed to me to be the best choice since LLVM has wide instruction semantics implementations, that more basic tools such as Xed do not offer.
If I get it correctly… you would rather re-implement such semantic yourself? Or have you got any other tool / idea I’m not aware of?
I don't regret using LLVM at all. I regret using the instruction decoding features in LLVM. if I could do it again (and someone younger, smarter, and better looking than me is) I would combine the LLVM IR with the XED instruction decoder. we would still emit the semantics of the instructions as LLVM, but we would use XED to figure out which instruction we were decoding.
What you did with McSema was really impressive and I’m glad discussing with one of the authors! I get the point you regret using LLVM, it seemed to me to be the best choice since LLVM has wide instruction semantics implementations, that more basic tools such as Xed do not offer. If I get it correctly… you would rather re-implement such semantic yourself? Or have you got any other tool / idea I’m not aware of?