Using OpenMP superscalar for parallelization of embedded and consumer applications

Abstract

Abstract-In the past years, research and industry have introduced several parallel programming models to simplify the development of parallel applications. A popular class among these models are task-based programming models which proclaim easeof-use, portability, and high performance. A novel model in this class, OpenMP Superscalar, combines advanced features such as automated runtime dependency resolution, while maintaining simple pragma-based programming for C/C++. OpenMP Superscalar has proven to be effective in leveraging parallelism in HPC workloads. Embedded and consumer applications, however, are currently still mainly parallelized using traditional threadbased programming models. In this work, we investigate how effective OpenMP Superscalar is for embedded and consumer applications in terms of usability and performance. To determine the usability of OmpSs, we show in detail how to implement complex parallelization strategies such as ones used in parallel H.264 decoding. To evaluate the performance we created a collection of ten embedded and consumer benchmarks parallelized in both OmpSs and Pthreads.

Parallelizing Parallel Programs

“…Our current implementation has rules capturing a variety of flavors of both Map and Reduce patterns, as reported in Table 1. Table 1 presents our results for multithreaded code from the Starbench suite [1], omitting two programs, bodytrack and h264dec, which follow a pattern (pipeline) that is out of the scope of our current analysis. We found all but one expected patterns, known by manual inspection and from the literature, and understand the reason for missing one map in kmeans, which will inform subsequent versions.…”

A Runtime Heuristic to Selectively Replicate Tasks for Application-Specific Reliability Targets

“…However, our selective task replication heuristics are applicable for other taskparallel dataflow platforms. Nevertheless, the performance of OmpSs+Nanos is on par with the highly optimized commercial and open source implementation of OpenMP [5], [6] and it has successfully served as a pilot platform to push dataflow task parallelism to OpenMP 4.0. In the case of the distributed OmpSs+MPI model, it combines dataflow execution with the message passing model providing significant performance benefits.…”

An Integrated Hardware-Software Approach to Task Graph Management

“…Listing 1 shows a simplified example of OmpSs programming, extracted from H.264 macroblock wavefront decoding [10]. The function decode() is called inside a nested loop, processing the elements of matrix X.…”