The main feature of the many-core accelerators such as GPU is their massively parallel architecture allowing them to speed up computations that involve matrix-based operations, which is a heart of many ML/DL implementations. Manufacturers often offer the possibility to enhance hardware configuration with many-core accelerators to improve machine/cluster performance as well as accelerated libraries, which provide highly optimized primitives, algorithms and functions to access the massively parallel power of GPUs (Table 2).

Other parallel programming libraries, which support computational speed-up, are:

• OpenMP: application programming interface (API) that supports multi-platform shared memory multiprocessing programming [OpenMP]. It consists of a set of compiler directives, library routines, and environment variables that influence run-time behaviour.
• Open MPI: open source, freely available implementation of the MPI specifications [Open MPI]. The Open MPI software achieves high performance and it is quite receptive to community input. MPI stands for the Message Passing Interface - a standardised API typically used for parallel and/or distributed computing. It is written by the MPI Forum, which is a large committee comprising of a cross-section between industry and research representatives.

An application built with the hybrid model of parallel programming can run on a computer cluster using both OpenMP and MPI, such that OpenMP is used for parallelism within a (multi-core) node while MPI is used for parallelism between nodes. More details about accelerators and accelerated computing will be available in Deliverable D4.1 Available Technologies for accelerators and HPC. In the recent years the accelerators have been successfully used in many areas e.g. text, image, sound processing and recognition, life simulations as ML/NN/DL applications. Applicable areas of DNNs are:

• Image and video processing: satellites images (such as fires, droughts, crops diseases, urban development), space (telescope images), biology image recognition (such as plant, cells, bacteria), medical image recognition (such as magnetic resonance imaging, computer tomography, roentgen images, sonography), automatic picture or audio annotations [Hafiane 2017];
• Speech and language: text processing and recognition, speech recognition, machine translation, natural language processing;
• Security: biometrics authentication (such as people, faces, gait), anomaly detection, intrusion detection;
• Business intelligence: insurance, financial markets, stock and exchange rate predictions;
• Robotics and videogames: autonomous navigation (such as car, drone, plane, submarine), videogames (such as Atari, Dota, Starcraft).

Return to Contemt