michael-emmi / c2s-ocaml Goto Github PK

Perhaps Boogie would be happier with an encoding without using maps, so that array theory can be avoided -- at least from the sequentialization. Instead of replacing global accesses G by G[k], we might instead introduce Boogie functions.

The Basic Idea

For each global G of type T, given K rounds, introduce the variables G.1, G.2, .., G.K of type T, and a function G(int,T,T,T,..,T) -- there should be K T's there. This function is defined such that G(i,g.0,g.1,..,g.K) = g.i. Each access to G is then replaced by G(k,G.1,G.2,..,G.K), where k is the current round index.

Will this help?

Unclear. But it certainly will avoid introducing array theory reasoning in Boogie's VCs. Other things to consider are:

• the number of assignments/equalities between global variables and saved/guessed values will multiply by a factor of K, thus increasing the size of the generated program by O(K).
• array theory might be present anyways, e.g., from SMACK translated code