Intel recently provided an update on the company’s software project, called One API. The project is meant to deliver a unified programming model to simplify application development for workloads, even when working across diverse computing architectures. Since the inception of the project, the breadth of compute has grown to include specialized accelerators. So Intel is now working to deliver software solutions that allow users to get the most out of their hardware. Intel is planning on releasing a developer beta and even more details on the project in the fourth quarter of 2019.
Basically, One API is meant to provide an efficient software programming model to deliver performance for those who are working with diverse, datacentric workloads. One API supports both direct programming and API programming. So it can deliver a unified language and libraries that provide full native code performance across a range of hardware. Here are some more details about the various features and options of this project.
Direct programming with One API
One API has a new direct programming language called Data Parallel C++. It’s an open, cross-industry alternative to single architecture proprietary languages. Based on C++, this programming language delivers parallel programming performance and productivity through a programming model that’s also familiar to developers. It also incorporates SYCL and includes language extensions that were developed using an open community process.
In addition to the direct programming options, One API also includes powerful libraries that span several workload domains that benefit from acceleration. These library functions are custom coded to work with each specific target architecture.
Analysis and debug tools
Intel is also working to deliver enhanced versions of its analysis and debug tools to support the DPC++ programming language and the range of different SVMS architectures included in One API. These tools are meant to help users make the experience even easier.
Photo credit: Freerange Stock