Institute of Information Science, Academia Sinica

User experience is an important differentiating feature for Intel future products. Perceptual computing (e.g., face recognition, gesture recognition and object tracking), usually demanding high computational power, plays a crucial role in enriching user experience. With the integrated graphic processors are becoming ubiquitous, new languages are developed to address GPGPU programmability, e.g., OpenCL, Compute Shader, MS C++ AMP. Writing applications using those languages, although achieving portability, does not guarantee to fully exploit the underneath HW features. The talk presents an on-going video acceleration library effort that provides C/C++ interface that numerous users are comfortable and familiar with while minimizing the runtime overhead between CPU and GPU and fully leveraging Intel Video Analytic Fixed Functions.