Source-to-Source Automatic Differentiation of OpenMP Parallel Loops

This article presents our work toward correct and efficient automatic differentiation of OpenMP parallel worksharing loops in forward reverse mode. Automatic is a method to obtain gradients numerical programs, which are crucial optimization, uncertainty quantification, machine learning. The computational cost compute common bottleneck practice. For applications that parallelized for multicore CPUs or GPUs using OpenMP, one also wishes the parallel. We propose framework reason about correctness generated derivative code, from we justify extension model. implement this model tool Tapenade present test cases differentiated following extended procedure. Performance programs mode better than sequential, although often scales worse input programs.