One API provides a unified programming model to simplify application development across different computing architectures

At Intel Software Technology Day held in London on June 19, Intel engineering experts introduced the latest progress of the software project "One API", which aims to provide a unified programming model to simplify application development across different computing architectures.

“The One API project will bring developers a set of tools that provide a unified programming model that simplifies development for workloads across multiple architectures,” said Bill Savage, Intel vice president and general manager of Computing Performance and Developer Products. “Intel’s broad computing portfolio now includes dedicated accelerators, and the software solutions we provide will help developers fully unleash the performance of the hardware.”

As data-centric workloads in today's world become increasingly diverse, the architectures for processing data are also becoming increasingly diverse. Intel's broad computing architecture covers scalar, vector, matrix, and spatial. At Intel, these architectures are often abbreviated as SVMs, and they require an efficient software programming tool to fully unleash their performance. The ease of use and high performance of One API meet this need, while eliminating the need to maintain different code bases, multiple programming languages, and different tools and workflows.

The Intel® OpenVINO™ toolkit is an example of how Intel currently provides this cross-architecture capability to customers. This software developer toolkit enables consistent, optimized deep learning inference capabilities across different SVMS architectures.

One API supports direct programming and API programming, and will provide a unified language and library that can provide full local code performance on different hardware including CPU, GPU, FPGA and AI accelerators.

• Direct Programming: One API includes a new direct programming language, Data Parallel C++ (DPC++), which is an open, cross-industry programming language that replaces single-architecture proprietary languages. Using a programming model familiar to developers, DPC++ delivers the efficiency and performance of parallel programming. DPC++ is based on C++, incorporates Kronos Group's SYCL*, and includes language extensions developed in an open community process.

• API-based programming: One API’s powerful library spans multiple workload areas that can benefit from acceleration. Library functions are custom-coded for each target architecture.

• Analysis and Debug Tools: Building on the leading analysis tools, Intel will provide enhanced analysis and debug tools to support DPC++ and the broad SVMS architecture.

Intel will release a One API developer beta in the fourth quarter of 2019 and disclose more details about the project.