Table of Contents
Intel Corp. right now introduced a new model of its oneAPI toolkit, which enables developers to develop purposes that can run on multiple kinds of processors.
Many companies’ technological innovation environments include several sorts of chips. An firm may possibly, for instance, operate its databases on central processing units and its synthetic intelligence purposes on graphics processing device cards. Some providers also use more specialized chips, such as discipline-programmable gate arrays.
An software that can run on a person form of chip is not necessarily suitable with one more. Typically, porting an application to a new processor architecture demands developers to make intensive code improvements. The more chips a workload will have to assistance, the more code changes have to be produced.
Intel’s oneAPI toolkit makes it simpler to develop applications that can operate on numerous styles of chips. According to the enterprise, oneAPI decreases the total of code that will have to be changed when an application is ported from 1 processor architecture to yet another. The end result is that developers can total computer software projects a lot quicker.
The latest version of oneAPI that Intel debuted currently attributes guidance for several of the company’s forthcoming details middle processors. It also provides enhanced compatibility with rival chipmakers’ silicon.
In accordance to Intel, the new oneAPI launch supports its future Sapphire Rapids collection of server CPUs. The company’s not too long ago released Xeon Max collection is supported as well. The latter products line features CPUs optimized for higher-performance computing systems, most notably supercomputers.
Developers can now also use oneAPI to establish apps for Intel’s impending line of facts center graphics playing cards. The products line is headlined by the Max Series, a chip with extra than 100 billion transistors. It contains 47 chiplets, or computing modules, optimized to run machine mastering software program.
Some purposes demand the potential to operate on GPUs from not just one but several suppliers. To relieve the enhancement of these types of workloads, Intel is extending oneAPI with a assortment of new software plugins. The plugins had been produced by Codeplay Software, an Edinburgh-based mostly business that Intel obtained earlier this calendar year.
Developers use a programming language recognized as SYCL to produce oneAPI apps. Code penned in the language can run on many varieties of chips. According to Intel, the new oneAPI plugins will make it simpler to write SYCL code that can operate on Nvidia Corp. and Innovative Micro Units Inc. graphics processing units.
Nvidia features a application toolkit named CUDA along with its GPUs. Firms use the toolkit to enhance the overall performance of their GPU-driven programs.
By default, CUDA-powered programs created for Nvidia chips aren’t compatible with other graphics cards. As a end result, porting such apps to one more chipmaker’s GPUs can entail a sizeable amount of money of work. To ease the job, Intel has updated oneAPI with characteristics that will make it much easier to change CUDA-driven program into SYCL code that supports numerous varieties of graphics cards.
Intel is also simplifying the progress of GPU-based mostly programs in other approaches. As component of today’s update, the corporation has added a element that can routinely execute roofline investigation of workloads. Roofline evaluation is a performance analysis system that allows builders find techniques of speeding up their software package.
“We’re seeing encouraging early software performance outcomes on our development techniques employing Intel Max Series GPU accelerators – applications constructed with Intel’s oneAPI compilers and libraries,” claimed Timothy Williams, the deputy director of the Argonne Nationwide Laboratory’s computational science division. “For management-course computational science, we value the gains of code portability from multivendor, multiarchitecture programming benchmarks.”
Along with the new GPU capabilities, the latest release of oneAPI also involves quite a few other enhancements. Intel has included improved help for OpenMP, an open-supply resource applied by builders to create programs that can operate on multiple processor cores at at the time. Intel also improved oneAPI’s oneMKL element, which would make it simpler to make programs that can accomplish complex mathematical operations.