StarPU: A Unified Platform for Task Scheduling on Heterogeneous Multicore Architectures

Abstract

Abstract. In the field of HPC, the current hardware trend is to design multiprocessor architectures that feature heterogeneous technologies such as specialized coprocessors (e.g., Cell/BE SPUs) or data-parallel accelerators (e.g., GPGPUs). Approaching the theoretical performance of these architectures is a complex issue. Indeed, substantial efforts have already been devoted to efficiently offload parts of the computations. However, designing an execution model that unifies all computing units and associated embedded memory remains a main challenge. We have thus designed STARPU, an original runtime system providing a highlevel, unified execution model tightly coupled with an expressive data management library. The main goal of STARPU is to provide numerical kernel designers with a convenient way to generate parallel tasks over heterogeneous hardware on the one hand, and easily develop and tune powerful scheduling algorithms on the other hand. We have developed several strategies that can be selected seamlessly at run time, and we have demonstrated their efficiency by analyzing the impact of those scheduling policies on several classical linear algebra algorithms that take advantage of multiple cores and GPUs at the same time. In addition to substantial improvements regarding execution times, we obtained consistent superlinear parallelism by actually exploiting the heterogeneous nature of the machine.

Improving Scalability and Maintenance of Software for High-Performance Scientific Computing by Combining MDE and Frameworks

“…Even though this feature is not yet implemented, we think that higher-order transformations can make this kind of parametric studies accessible. In the same spirit, projects such as StarPU [17] require the algorithm to be implemented in different languages. StarPU is a unified runtime system that offers support for heterogeneous architectures (CPU, GPUs, IBM Cell) by selecting at runtime the more relevant implementation.…”

High-throughput Analysis of Large Microscopy Image Datasets on CPU-GPU Cluster Platforms

“…The appropriate utilization of hybrid systems, however, typically requires complex software instruments to deal with a number of peculiar aspects of the different processors available. This challenge has motivated a number of languages and runtime frameworks [15], [14], [16], [18], [19], [20], [21], [22], [23], [24], [25], [26], specialize libraries [4], and compiler techniques [27]. …”

A Static Task Partitioning Approach for Heterogeneous Systems Using OpenCL

“…StarPU [4] is a framework for programming on heterogeneous systems. The user provides multiple versions of each task and the run-time schedules them to the devices.…”