Closed 2over12 closed 7 months ago
Hello, I'm so sorry for the late reply, there was the PLDI panic and then it slipped my mind
The name bi_unroll_depth is misleading indeed, it's actually implemented by capping the allowed number of branches. This is not the best heuristics for bounding bi-abduction, but it is sound. Implementing another heuristics would require deeper reworking that is not at the top of our priority list for now :(
Was messing around with ACT on a simple example:
With default options this dumps no specs:
esy x gillian-c act /tmp/t.c --specs-to-stdout -l verbose --reporters="file"
I realized this is because bi-unroll is also tied to the max branching factor which seems odd and unexpected to me: https://github.com/GillianPlatform/Gillian/blob/500b9069a64bfd5175f6ff8de522db29112ef72d/GillianCore/command_line/act_console.ml#L146
Is there any reason not to just have its own parameter?