Tuning Data and Control Structures for Incremental Computation
General purpose incremental computation requires a lot of overhead in memory and time. To address these limitations, programmers often organize functions and data so that dependency tracking and memoization act on groups rather than individual items. I design data structures and higher-order functions in the Rust language with support for tuning these group sizes. I use the most recent, the Giraz, to parse and interpret a reverse-polish-calculator. The Giraz has high initial runtime compared to a native vector-based implementation, but makes up for it with fast change propagation after 10 changes.
Thu 22 Jun
|15:50 - 16:20|
|16:25 - 16:55|
Mirko KöhlerTechnical University of Darmstadt, Philipp HallerKTH Royal Institute of Technology, Sebastian ErdwegTU Delft, Mira MeziniTU Darmstadt, Guido SalvaneschiTU DarmstadtFile Attached
|17:00 - 17:30|
Kyle HeadleyUniversity of Colorado BoulderFile Attached