Maybe we should just work off of the assumption that there's no relevant uncomputable physics, because if there were, we should probably give up our endeavors anyways, unless we knew how to model an uncomputable reality within a computable AGI-program. As Schmidhuber ever so aptly wrote on his homepage:

The distribution should at least be computable in the limit. That is, there should exist a program that takes as an input any beginning of the universe history as well as a next possible event, and produces an output converging on the conditional probability of the event. If there were no such program we could not even formally specify our universe, leave alone writing reasonable scientific papers about it.