I've seen many papers on scheduling components or tasks once a DAG for the program is known, either by user-input or by domain restriction (i.e. all cross shaped 5-pt stencil codes have a known DAG).
Is there an algorithm for taking code (or Abstract syntax trees) from a language, say Python, and generating a DAG for it? I've been searching for a paper related to this for a long time and haven't found anything on google scholar.
Intuitively, I'd say that it's not possible, since some languages (such as Python) can change just about everything at run-time, and thus to determine the behavior and dependencies of a given program you would have to run it first for every possible set of inputs.
EDIT: I'm using DAG in this sense as the Directed Acyclic Graph for the dependencies (i.e. which parts depend on which other parts)