Intel oneAPI will arrive in December, the development platform to unify CPUs, FPGAs, GPUs and other accelerators of the company

Source: Geeknetic added 11th Nov 2020

  • intel-oneapi-will-arrive-in-december,-the-development-platform-to-unify-cpus,-fpgas,-gpus-and-other-accelerators-of-the-company

by Antonio Delgado 3 hours ago

In December of this year Intel will publish oneAPI in its version 1.0, a new development software platform, announced last year, with which they want to unify different accelerators such as CPUs, GPUs, FGAs and other chips for more specific tasks. This release is part of the company’s new strategy and its approach towards what they call XPUs (CPUs, GPUS, FGAs and other accelerators) from their previous strategy of focusing on CPUS.

It will consist of a set of basic tools, called I ntel oneAPI Base Toolkit, where the core of the software is included, including different compilers, compatibility, debug and analysis tools, as well as high-performance libraries and other tools with which to get the most out of XPUs without the need for a software specific development for each one. All this under the umbrella of the Data Parallel C ++ programming language.

oneAPI will allow to port applications for other platforms such as CUDA in a simple way

Although the oneAPI language will be DPC ++, the compatibility tools will allow converting from other languages ​​and environments.

For example, an application like EasyWave, written in CUDA for NVIDIA GPU accelerators, will be able to port to oneAPI quickly and will be optimized to work on all types of Intel architectures, from CPUs, to GPUs and FPGAs or even third-party architectures.

The OneAPI Toolkits include a tool designed for artificial intelligence (AI Analytics Toolkit), designed for researchers and developers of and AI. In addition, OpenVINO Toolkit will allow to take advantage of the acceleration in deep learning environments in the process of inference and computerized vision.

Additionally, the company will launch Toolkits specific to different environments, such as the oneAPI HPC ToolKit intended for high-performance computers and servers, with optimizations for environments and languages ​​like Fortran, OpenMP, MPI and the like. Another Toolkit available will be oneAPI IoT, designed for Internet of Things devices and environments ranging from smart homes up to aerospace or medical industry.

Thought for graphics we will have the oneAPI Rendering Toolkit, with an orientation towards display systems, for both leisure and industry, entertainment, research, etc.

Specifically, oneAPI will support, within the Windows and Linux ecosystems, much of Intel’s architectures, from Intel Core to Atom and Xeon, in addition to its Gen 9 graphics, Iris Xe and variants and its Stratix and Arria FPGAs:

  • CPUs – Intel® Xeon®, Core ™, Atom, and supported processors
  • GPUs – Intel® Processor Graphics Gen9, Intel Iris Xe MAX graphics, Intel Xe -HP
  • FPGAs – Intel® Arria®, Intel® Stratix 11 FPGAs

Support for oneAPI will be adopted by many companies and institutions.

End of Article. Tell us something in the Comments or come to our Forum!