Abstract In the era of heterogeneous computing, a new paradigm called accelerator level parallelism (ALP) has emerged. In ALP, accelerators are used concurrently to provide unprecedented levels of performance and energy efficiency. To reach that there are many problems to be solved, one of the most challenging being co-execution. In this paper, we present a new scheduling framework called POAS, a general method for providing co-execution to applications. Our proposal consists of four steps: predict, optimize, adapt and schedule. With POAS, an unseen application can be executed concurrently in ALP with little effort. We evaluate POAS on a heterogeneous environment consisting of CPUs, GPUs (CUDA cores), and XPUs (Tensor cores) on two different fields, namely linear algebra (matrix multiplication benchmark) and deep learning (convolution benchmark). Our experiments prove that POAS provides excellent performance and completes the tasks within a time very close to the optimal time for the hardware and applications used, with a negligible execution time overhead. Moreover, the POAS predictor performed exceptionally well, achieving very low RMSE values for both use cases. Therefore, POAS can be a valuable tool for fully exploiting ALP and improving overall performance over offloading in heterogeneous settings.

Dedicated tensor accelerators demonstrate the importance of linear algebra in modern applications. Such accelerators have the potential for impressive performance gains, but require programmers to rewrite code using vendor APIs - a barrier to wider scale adoption. Recent work overcomes this by matching and replacing patterns within code, but such approaches are fragile and fail to cope with the diversity of real-world codes.

Abstract Different technologies and approaches exist to work around the performance portability problem. Companies and academia work together to find a way to preserve performance across heterogeneous hardware using a unified language, one language to rule them all. Intel's oneAPI appears with this idea in mind. In this article, we try the new Intel solution to approach heterogeneous programming, choosing machine learning as our case study. More precisely, we choose Caffe, a machine learning framework that was created six years ago. Nevertheless, how would it be to make Caffe again from the beginning, using a fresh new technology like oneAPI? In terms of not only the ease of programming‐because only one source code would be needed to deploy Caffe to CPUs, GPUs, FPGAs, and accelerators (platforms that oneAPI currently supports)‐but also performance, where oneAPI may be capable of taking advantage of specific hardware automatically. Is Intel's oneAPI ready to take the leap?

Abstract This paper presents HDNN, a proof-of-concept MLIR dialect for cross-platform computing specialized in deep neural networks. As target devices, HDNN supports CPUs, GPUs and TPUs. In this paper, we provide a comprehensive description of the HDNN dialect, outlining how this novel approach aims to solve the $$P^3$$ P 3 problem of parallel programming (portability, productivity, and performance). An HDNN program is device-agnostic, i.e., only the device specifier has to be changed to run a given workload in one device or another. Moreover, HDNN has been designed to be a domain-specific language, which ultimately helps programming productivity. Finally, HDNN relies on optimized libraries for heavy, performance-critical workloads. HDNN has been evaluated against other state-of-the-art machine learning frameworks on all the hardware platforms achieving excellent performance. We conclude that the ideas and concepts used in HDNN can be crucial for designing future generation compilers and programming languages to overcome the challenges of the forthcoming heterogeneous computing era.

Ant Colony Optimization (ACO) is a population-based meta-heuristic inspired by the social behavior of ants. It is successfully applied in solving many NP-hard problems, such as the Traveling Salesman Problem (TSP). Large-sized instances pose two memory problems to the ACOTSP algorithm: the memory size and the memory bandwidth. This work has focused on developing ACOTSP-MF, a new ACOTSP algorithm proposed to adequately manage the memory issues that arise while solving large TSP instances. ACOTSP-MF uses the nearest neighbor list, introducing a novel class of cities, the backup cities, while grouping cities into three classes: the nearest neighbor cities, the backup cities, and the rest of the cities (the majority). ACOTSP-MF also modifies how the base ACOTSP carries out the tour construction and pheromone update phases depending on the group to which a city belongs. This way, ACOTSP-MF reduces both the memory requirements of its data structures (from O(n∗n) to O(n)), and the memory bandwidth needs (thanks to better exploitation of the memory data locality). In this paper, we have carried out an in-depth analysis of ACOTSP-MF performance for medium and large TSP instances, covering vectorization and scalability issues and showing its main bottlenecks. For medium-size instances, the paper reports speedup factors of 20-500X for the rl11849 instance compared to the base ACOTSP version. ACOTSP-MF is intended and especially adequate for large-size instances. In this context, the paper reports excellent execution time for the Tour Construction phase, with less than 500 ms per iteration for the earring-200k instance. Finally, a study about the solution quality of ACOTSP-MF has been included, showing that ACOTSP-MF paired with local search offers high solution quality (within 2% of the best-known solution).