Re:Swiftc – SIL and its not-so-basic blocks
by Rusky on Monday November 02, 2015 @08:03PM
(#10496660)
I've seen another IR that replaced phi nodes with BB arguments, but they also replaced return with continuations, leading to a kind of best-of-both-worlds cross between SSA and CPS. They were able to implement a bunch of interesting optimizations on closures, in an imperative language.
http://compilers.cs.uni-saarland.de/papers/lkh15_cgo.pdf