Chief Video IP Solutions Architect
Johan started working at Xilinx almost 5 years ago, where he initially built the Video IP solution portfolio for all Xilinx video applications. This spans from video connectivity (HDMI, DP, SDI) to Video Processing (scalers, CSC, Deinterlacers, etc) and Video Over IP solutions. This portfolio which was built up from the ground is now mature and part of Xilinx IP catalogue. In the recent 1.5 years, Johan focused on enabling FPGAs for compute acceleration for video in Data centers/server, by architecting and developing the Video Acceleration solution from conception to production, with focus on Video Transcoding. As part of this, he drove the development of video transcoding in cloud/server & Data Center applications (HEVC, h.264, VP9) and associated compute acceleration. This portfolio is developed in a plug-play manner, where the FPGA is abstracted from the software developer and application. For example by enabling FFMPG plug-ins for codec acceleration. More recently, he is working on enhancing this portfolio with Machine Learning technologies Before Johan joined Xilinx, he worked as System/SoC/Video Architect Fellow for semiconductor companies like NXP and STmicro, and as VP of Architecture at Trident Microsystems where he was responsible for architecting the consumer SoCs for TV and Settopbox markets.
Live video traffic is growing faster than any other video traffic type and China’s video operators are witnessing this first hand. The volume of traffic shows no signs of slowing down and this is putting pressure on existing infrastructure and associated financial models that providers have relied on since live streaming’s inception. The industry is desperate for a new approach that will enable lower bandwidth requirements, reduced infrastructure costs while simultaneously maintaining the agility that software has provided. Field Programmable Gate Arrays or FPGAs can address these needs and have most recently gained popularity in leading Live Streaming applications. Hardware acceleration for live streaming is very attractive but companies has existing implementations and software that they cannot disrupt or change easily. During this talk Xilinx will highlight will outline how FPGAs can be used in existing applications and future requirements like Video + Machine Learning. We Xilinx has integrated with FFmpeg and engineers can utilize FPGAs in their existing networks without having to make significant changes to their software infrastructure. We will also demonstrate how Machine Learning can be integrated In to FFmpeg enabling acceleration all through command line interface.