Open source syclneural networkLibrary for powerVROptimized to make it easier for developers to migrate existing code by working with codeplay.
London, UK and Santa Clara, USA, October 23, 2019 – imaginationTeChnologies announced that thanks to the new and optimized open-source sycl neural network library, usingTensorFlowDevelopers will be able to directly target powervrGraphicalProcessor (GPU）Development. Its first version will be commercially available in 2019.
TensorFLow’s sycl version supports a large number of artificial intelligence（AI）Operation (as shown in Figure 1), and users are easy to customize on demand, which means that developers can run those instantly available high-performance networks on powervr by using the latest neural network or the AI technology researched by them. Because tensorflow sycl support is both open source and open standard based, it is an ideal solution for developers who want to accelerate the latest AI technology on low-power devices.
Figure 1 – tensorflow operations supported by sycl
Sycl is an open standard solution that can replace CUDA architecture without paying royalties. It breaks the barriers between ecosystems, so as to provide more freedom for developers to use standard C + + to write code, release the performance advantages of GPU hardware and ensure the code portability.
In addition, codeplay’s sycl library enables applications to seamlessly utilize applications optimized for imgdnn powervrInterface（API) Imgdnn is a special neural network graphics compilation Library of imagination, which can help developers get the highest performance from powervr GPU and neural network accelerator (NNA).
The main markets for this new and expanded ecosystem include cars, data centers and smart cameras.
ImaginaTIon Technologies产品管理高级总监Neal Fors“Working within an open standards framework reassures developers that their code is not out of date or needs to be rewritten,” says E Through sycl, you can access the widely used powervr GPU, and developers can easily obtain powerful computing resources under tensorflow. “
Andrew Richards, CEO of codeplay, said: “now we see a huge growth in market applications of sycl standards.RezaRenesas has enabled sycl on its r-car AI platform, and nowIntelSycl is also included in its one API. Programming high-performance accelerators, such as powervr GPUs, through increased standardization will enable AI software developers to bring advanced intelligence into everything, including from micro to low-powerBatteryDrive device to large superComputer。”
Sycl is based on the concept and efficiency of Khronos OpenCL ™. Sycl libraries optimized for powervr, including sycl-dnn, sycl-blas and eigen, will be available on GitHub. The tensorflow branch with extended sycl support is available from the GitHub of codeplay.