I have converted the paper to the PLDI format. Then I fixed a bunch of spacing issues (severe overflow and serious space under-use) to get a first approximation. We’re basically at 16 pages (16.5 but the draft new introduction is about .4 pages). We need to be at 12 pages.

I have no clue how to remove 4 whole pages of content. There are still a bunch of whitespace things to deal with, but half of them will consume space (I did not fix all overflows).

With quite a bit of effort, I think we could cut out about a page and still have this paper convey what it currently does. Beyond that, I just don’t see it.

Ok, I fixed more whitespace issues. Miraculously, it didn’t make any real change to the length. [Some are a bit hacky, but I wanted to get an estimate]. There’s a few still to do (and the diagram too, but I have ideas on how that can be done, if it is necessary).

Title brainstorming: - From human-friendly probabilistic programs to machine-efficient code - Bridging the gap between high-level inference algorithms and efficient code - Probabilistic cake: high-level inference algorithms and efficient code too! - Efficiency at both ends: the case of probabilistic languages - Domain-specific optimizations FTW: the case of probabilistic languages - Algebra + Analysis + Probability + Compiler = fast probabilistic programs