Benchmarking, parallel array reduction and interval arithmetic in Delite

During the past week Adam has been gathering performance measurements for the Acumen enclosure interpreter. He has been focusing on specific simulator parameters (e.g. the minimum ODE solver– and event localisation time steps) and the sensitivity of the end-time precision with respect to these. In order to make the measurements reproducible, Adam has adapted the benchmarking shell scripts that Kevin prepared previously. Further, Adam has been working on the assignments for the Heterogenous Parallel Programming course on Coursera, which this week involved implementing parallel array reduction in CUDA. Adam also prepared a project description for the project part of the Chalmers Scala course he attended in December. The project will aim to build an interval arithmetic library on top of the Delite framework, with a code generator for either OpenCL or CUDA. Initially, Adam plans to approach the implementation of the library by developing a sequential code generator for Scala.