This is a nice VM, it has an interesting stack items model, better separation
of elements from stack implementation, simpler stack implementation and a bit
nicer opcode implementation mechanism. At the same time it lacks so many
features and is so differing from our current VM (that is closely tied to the
compiler) that it makes its code very hard to reuse for master's VM
improvement.
Also, some differences are non-obvious to judge in terms of better or
worse. Stack item model seems to be more extensible, but at the same time we
know very well what kind of stack items we have and this doesn't change a
lot. Slice-based stack is simple but it's hard to say which one would perform
better for real-world smart contracts (it has different tradeoffs in
operations complexity).
Based on that, I'm dropping it. Some ideas will be reused during VM
refactoring, but no more than that. Refs. #307.
d4d1c79546