ComputeCpp, Codeplay's implementation of the open standard SYCL, enables you to integrate parallel computing into your application and accelerate your code using OpenCL devices such as GPUs. Applications that require a large number of common operations can make huge performance improvements by running the operations in parallel on OpenCL devices. For example, the neural networks used in machine learning perform huge numbers of matrix calculations and ComputeCpp can be used to run these operations in parallel, vastly increasing performance and reducing the power used by the application.
With ComputeCpp and SYCL you can write code once and execute on a range of OpenCL enabled devices reducing your development effort. Develop with standard C++ and the SYCL open standard, re-using your existing C++ libraries. ComputeCpp is also building support for C++17 Parallel STL enabling parallelized library functions to run on accelerated processors. ComputeCpp also works with a number of frameworks including ParallelSTL and VisionCpp.
Portable Parallel Computing Applications
OpenCL devices such as GPUs can be used to accelerate applications by running operations in parallel. By implementing ComputeCpp using the SYCL open standard, developers can write software with C++ single source and run their code using parallel computing across a range of OpenCL devices.
Using TensorFlow with ComputeCpp
Machine learning framework TensorFlow requires large amounts of vector and matrix operations. Performance and power consumption can be vastly improved by using parallel computing. ComputeCpp enables developers to target OpenCL devices such as GPUs using modern C++ code.
Artificial Intelligence Applications
Performing complex image processing operations can be accelerated using parallel computing. ComputeCpp enables high-level programmability for custom vision processors, enabling additional custom features on top of existing optimized hardware functions.
Complex Mathematical Applications
The Eigen library is one of the most popular C++ libraries for linear algebra, matrix and vector operations and related algorithms. Eigen is integrated with ComputeCpp enabling developers to run these operations on OpenCL devices. By taking advantage of these parallel architectures, applications can be accelerated.
University of Münster
WIGNER Fizikai Kutatóközpont
The French Aerospace Lab
Shaping a Scalable Future
University of the West of Scotland