In December 2020 Intel Software officially launched Intel oneAPI. We have had questions about what oneAPI is and what that means for Intel Parallel Studio and Intel System Studio customers – how do they transition? In this FAQ-style blog series, we will delve into all this in more detail for you.
In the first part of this series, we will cover the very basics of what oneAPI is.
What is oneAPI?
oneAPI is an industry initiative creating an open, standards-based, cross-architecture programming model to simplify development for a wide range of data-centric workloads across a variety of architectures (CPU, GPU, FPGA, and other accelerators). It includes a cross-architecture language Data Parallel C++ (based on ISO C++ and Khronos Group’s SYCL), advanced libraries, and community extensions.
What’s included in the oneAPI Toolkits?
Intel has created a product implementation of oneAPI with the Intel oneAPI Toolkits. These help developers efficiently build, analyse, and optimise high-performance, cross-architecture applications for Intel CPUs, GPUs and FPGAs.
Note: The tools that comprised Intel Parallel Studio XE and Intel System Studio are now integrated into Intel’s oneAPI Products. The Intel oneAPI Toolkits are upward compatible supersets of the features in the previous studio products, so the studio products will no longer be offered to new customers in 2021.
- Intel oneAPI Base Toolkit
- Intel oneAPI HPC Toolkit
- Intel oneAPI IoT Toolkit
- Intel oneAPI Rendering Toolkit (please contact us for more information)
Here is a breakdown of what’s included in three of the main toolkits:
Find out more about these toolkits in part 2 of this blog series.
Will you have to port or rewrite existing code to take advantage of oneAPI?
No, oneAPI coexists with existing languages and programming models such as C, C++, Fortran, Python, MPI and OpenMP. However, if you have an application written in CUDA, some parts may need to be rewritten. Although, the Intel DPC++ Compatibility Tool will help CUDA developers migrate to oneAPI. Note, you will have to manually edit the output of the tool to address code that could not be migrated.
Which CPU, GPU, FPGA and specialised accelerators are supported?
Intel oneAPI currently supports Intel CPUs (Intel Xeon, Core, Atom), Intel Arria and Stratix 10 FPGAs, as well as the Intel Gen9 and Gen11 Processor Graphics as a proxy development platform for future Intel discrete data center GPUs. More Intel accelerator architectures will be added over time.
Will the oneAPI specification element be open sourced?
Many of the components and libraries are or will soon be open sourced, including the oneAPI DPC++ Compiler.
Find out more
Part 2 of this blog series gives more detail on the Intel oneAPI Toolkits – what’s included, how the licensing works and so on. Read Part 2 now.
If you have any initial questions, please get in touch. Grey Matter is honoured to be an Intel Software Elite Reseller. We work very closely with Intel and know the ins and outs of how the licensing works. We can help you navigate the different options available to ensure that you are using the right tools for your development needs. Call +44 (0) 1364 655 123 or email firstname.lastname@example.org.
oneAPI technical webinar series
We’re hosting a 3-part Intel oneAPI webinar series alongside Intel Software and Bayncore, 27-29 April.
Session 1: Porting Intel Parallel Studio code to Intel oneAPI – how to make the right choice.
Session 2: Introducing heterogeneity to legacy applications using Intel oneAPI
Session 3: Optimised deployment of convolutional networks (CNN) using the Intel Distribution of OpenVinoTM
Post updated 07/04/2021