Wednesday, August 19, 2026

August 19, 2026 slides

 https://drive.google.com/file/d/16JlfN16W131zwgTw9GZfeaaNxF_kNxhq/view?usp=sharing


Design optimization is a methodology for picking some feasible arrangement of things so as to maximize a utility: naturally this involves mathematical optimization. Software systems can often be cast as discrete constructions. So then, to what extent can whole software components be "solved" for? Also, just as discrete structures are rich in complexity, they can also be intractable to effective procedures. What regularity conditions in these problems can be leveraged? I'll examine a collection of problems and focus on one in particular: matrix algebra compilation.