Debugger for Multi-level Hybrid Parallel Programs on Heterogeneous Accelerator Cluster Architectures – Survey and Challenges
نویسنده
چکیده
The need to debug hybrid parallel programs on heterogeneous accelerator clusters opens a new set of challenges for concurrently managing the processes and threads at node and accelerator levels. Currently, there exist open source debuggers for traditional HPC clusters, which support debugging of multi-node parallel programs. At present, debugging at the accelerator level is handled through language-specific debuggers provided by the device manufacturers. It is desired that a standards based debugger for hybrid parallel programs should support multiplelanguages, multiple devices, ease of use, scalability, and portability. However, currently there is no open-source standardized debugger for hybrid parallel programs on heterogeneous multi-accelerator clusters. The authors in this paper survey the existing debugger solutions for multi-level hybrid parallel programs. The authors in this paper bring out the challenges involved in developing a standardized open source debugger based on GDB, for heterogeneous accelerator clusters and present the features of a debugger on heterogeneous multiaccelerator clusters.
منابع مشابه
Accelerator Exoskeleton
To maximize performance and power efficiency, future multi-core architectures may be heterogeneous, incorporating some accelerator cores alongside the IA cores. Accelerator Exoskeletons provide a shared virtual memory heterogeneous multi-threaded programming paradigm for these accelerators using novel CPU instruction set extensions and software tool chains with an Intel Architecture (IA) look-n...
متن کاملIDLI: An Interactive Message Debugger for Parallel Programs Using LAM-MPI
Many complex and computation intensive problems can be solved efficiently using parallel programs on a network of processors. One of the most widely used software platforms for such cluster computing is LAMMPI. To aid development of robust parallel programs using LAM-MPI we need efficient debugging tools. However, the challenges in debugging parallel programs are unique and different from those...
متن کاملExperiences with Achieving Portability across Heterogeneous Architectures
The increasing computational needs of parallel applications inevitably require portability across popular parallel architectures, which are becoming heterogeneous. The lack of a common parallel framework results in divergent code bases, difficulty in porting, higher maintenance cost, and, thus difficulty achieving optimal performance on target architectures. Our paper examines two representativ...
متن کاملExploring the Parallel Programming Design Space of Proximate, a Multi-Tile Programmable Accelerator
The slowing of Moores law and Dennard scaling is limiting the performance improvements of single core processors. Increasing clock frequency any farther will lead to high leakage current and infeasible power consumption. Over the past decade, focus has been shifted to multi-core processors to increase throughput by having multiple cores to target different types of parallelism instruction (ILP)...
متن کاملThe Distributed Application Debugger
Developing parallel programs which run on distributed computer clusters introduces additional challenges to those present in traditional sequential programs. Debugging parallel programs requires not only inspecting the sequential code executing on each node but also tracking the flow of messages being passed between them in order to infer where the source of a bug actually lies. This thesis foc...
متن کامل