Auto-tuned OpenCL kernel co-execution in OmpSs for heterogeneous systems

Abstract

The emergence of heterogeneous systems has been very notable recently. Still their programming is a complex task. The co-execution of a single OpenCL kernel on several devices is a challenging endeavour, requiring considering the different computing capabilities of the devices and application behaviour. OmpSs is a framework for task based parallel applications, that does not support coexecution between several devices. This paper presents an extension of OmpSs that solves two main issues. First, the automatic distribution of datasets and the management of device memory address spaces. Second, the implementation of a set of load balancing algorithms to adapt to the particularities of applications and systems. All this is accomplished with negligible impact on the programming. Experimental results reveal that the use of all the devices in the system is beneficial in terms performance and energy consumption. Also, the Auto-Tune algorithm gives the best overall results without requiring manual parameter tuning. 45 diverse nature of kernels prevents the success of a single data-division strategy in maximising the performance and efficiency of a heterogeneous system. Aside from kernel behaviour, the other key factor for load distribution is the configuration of the heterogeneous system. For the load to be well balanced, each device must get the right amount of work, adapted to the capabilities of 50 the device itself. Therefore, a work distribution that has been hand-tuned for a given system is likely to underperform on a different one. The OmpSs programming model presents a change of paradigm in many ways. It provides support for task parallelism due to its benefits in terms of performance, cross-platform flexibility and reduction of data motion [9]. The 55 programmer divides the code in interrelating tasks and OmpSs essentially orchestrates their parallel execution maintaining their control and data dependences. To that end, OmpSs uses the information supplied by the programmer, via code annotations with pragmas, to determine at run-time which parts of the code can be run in parallel. It enhances OpenMP with support for irregular 60 and asynchronous parallelism, as well as support for heterogeneous architec-

OptCL: A Middleware to Optimise Performance for High Performance Domain-Specific Languages on Heterogeneous Platforms

“…Other works [14,18,21] proposed novel adaptive scheduling mechanisms to partition the workload at the data level aiming at achieving load-balancing or power-saving. Many frameworks built based on these designs [8,16,19] can operate directly on OpenCL kernels. They typically start with a small portion of workload assigned to CPUs and the remainder to the accelerators (or vice versa).…”

OpenABLext: An automatic code generation framework for agent‐based simulations on CPU‐GPU‐FPGA heterogeneous platforms

“…To partition generic programs dynamically at runtime, several authors have shown that partitioning on a data level is a viable option for both regular and irregular problems. [26][27][28][29] Some works [30][31][32] tackle the problem of accelerating numerical algorithms such as matrix multiplication and the fast Fourier transform on heterogeneous systems using data partitioning approaches. In ABSs, however, we observe a strong locality of dependencies, as agents primarily interact with nearby agents, that is, their neighbors.…”

“…The runtime behavior of ABS not only potentially varies based on input parameters but can also substantially change over the course of a simulation run, requiring regular retraining of machine‐learning models to achieve good performance. To partition generic programs dynamically at runtime, several authors have shown that partitioning on a data level is a viable option for both regular and irregular problems 26‐29 . Some works 30‐32 tackle the problem of accelerating numerical algorithms such as matrix multiplication and the fast Fourier transform on heterogeneous systems using data partitioning approaches.…”