Thomas Haas

Yes, this means the are no possible targets to read from. It could be a bug in the analysis but also a correct result. For example, this warning also appeared...

> > Yes, this means the are no possible targets to read from. It could be a bug in the analysis but also a correct result. For example, this warning...

> The alias analysis produces more precise results when flattening is used (and thus it needs to handle more complex expressions; I would have expected that it would loose precision...

Based on your initial description and my understanding of the algo, let me suggest another (hopefully better) description: ``` The analysis constructs a directed inclusion graph over expressions of the...

> Your description looks fine to me. Assignments can also have multi-linear expressions. Yes, that's why I have the part about `k*x` being understood as a vector. > The underlying...

After looking into the code I found the different use-cases of `Offset` very confusing, especially why `Offset` is used instead of `Variable` (i.e., the offset itself acts as a variable)....

I feel like your newest change has just shifted the variable vs. edge problem, no? Previously, you used edges that sometimes represented variables. Now you use variables that sometimes represent...

> I fixed a bug in the analysis, where the communication rule was not correctly implemented. > > The rule states that from (X+o1←storesW+o2), (W⊇A+o3), (A+o4⊆R), (R+o5→loadsY) and ((o2+o3)∩(o4+o5)≠∅), you...

EDIT: I was wrong. The code uses `join` for sequential composition and `addInto` for parallel composition. The latter does not merge edges which contradicts the class description/comment, so the description...

Did you intentionally revert your recent changes? If so, you also reverted the changes to the handling of unary expressions, so they are still wrong.