oneAPI
oneAPI is a cross-industry, open, standards-based unified programming model that delivers a common developer experience across accelerator architecture - for faster application performance, more productivity, and greater innovation.
phone_forwardedContact Us
Our oneAPI Contributions
As part of the open source DPC++ LLVM project, Codeplay® has developed a way for developers using this framework to target NVIDIA® GPUs. This implementation uses the native CUDA® interface to enable both portability and performance for developers. This interface is available for both DPC++ and the MKL-BLAS linear algebra maths library which is part of the oneMKL maths framework. This work means that software developers can target all NVIDIA GPUs using SYCL™ code without needing to use any intermediate layers, exposing the full performance of the underlying hardware.
Find out how to use DPC++ for CUDA on our website.
menu Get Started with DPC++ for CUDASoftware Building Blocks For Your Processor
Codeplay can deliver a fully integrated solution for your processor that enables software developers to take full advantage of the benefits that your hardware design delivers. By providing an interface that supports the oneAPI industry standard, developers can accelerate the key parts of their software using the most common development frameworks, libraries and languages.
Custom vs. Standard Implementation
oneAPI Level 0
The oneAPI Level 0 specification has been defined to enable low-level direct-to-metal interfaces that enables the other components of oneAPI. The definition can be found on the oneAPI website and provides a flexible industry standard way to enable what software developers need to make the most of your processor.
Implementing oneAPI Level 0 means that you get all the higher-level oneAPI compiler tools running on top of this, including SYCL and DPC++, and libraries.
oneAPI DPC++
DPC++ provides the features needed to define data parallel functions and launch them on processors using standard C++ code. DPC++ is an implementation of the SYCL open standard and builds on oneAPI Level 0, enabling software developers to use standard programming languages and models to run complex applications using parallel programming.
Codeplay's ComputeCpp implementation can also be adapted to deliver these benefits for your processor and support the oneAPI interface.
oneAPI Math Kernel Library
oneMKL is a set of highly optimized maths libraries that are important for a wide range of software applications performing HPC and artificial intelligence. The oneAPI standard defines the oneMKL common interface for functions that provide BLAS, FFT, and vector mathematics.
Codeplay can optimize and fine tune these to deliver the best possible performance for your processor.
oneAPI Deep Neural Network Library
The Deep Neural Library oneDNN interface defines a performance library for common operations required to build complex machine learning applications.
Codeplay can optimize and fine tune these to deliver the best possible performance for your processor.