During the past week, Adam has finished a first reading of A Lattice-Theoretical Approach to Deterministic Parallelism with Shared State by Lindsey Kuper and Ryan R. Newton which introduces LVars as a novel building block for deterministic parallel programming. LVars are variables which support put operations of monotonically increasing values (with respect to a given domain) and can be thought of as a generalisation of IVars (single-assignment variables). Further, Adam and Jan have been working on adapting the previously constructed symbolic differentiation implementation (more specifically, its "smart constructors") to improve the execution time of a bipedal robot model written in Mathematica and Matlab, with some success – a first attempt which is still missing several optimisations gave a 10% performance improvement. During the coming week Adam plans to focus on creating PBTs of the most critical parts of safesolver, and to begin looking in to the LMS framework Delite.
