December 18, 015, Beijing, China] Following the launch of the second generation of distributed computing concepts and the establishment of the Alliance as a milestone in the global computing arena in September this year, to accelerate the further spread of second-generation distributed computing And landing applications, OpenHW open source hardware community (http://www.openhw.org), IBM and Xilinx jointly announced the first IBM-Xilinx Heterogeneous Computing Competition officially launched today, the most innovative for the national engineering and academic research field Heterogeneous computing solutions and provide strong technical support for the implementation of these solutions.
Academic experts from Tsinghua University, Shanghai Jiaotong University, Fudan University, Tianjin University and Beijing Jiaotong University, and enterprise experts from IBM and Xilinx will form a strong expert review team to focus on and support each team from idea to realization. At the same time, in order to encourage the participating teams to actively participate in the competition, the organizing committee specially set up a prize prize. The winners will also receive core FPGA chips donated by Xilinx. In addition, the winning entries will also have the opportunity to be presented at the four major FPGA international academic conferences around the world.
The outbreak of new application workloads such as cloud computing, big data, and cognitive computing has ushered in the second generation of distributed computing with the hybrid heterogeneous computing of general-purpose processors and application accelerators as the main features. In order to promote the application of the second generation of distributed computing, IBM and Xilinx have launched a series of strategic cooperation around the world, such as jointly developing related acceleration infrastructure, software and middleware technologies and solutions. An important measure for the popularization of the second generation of distributed computing models.
The first IBM-Xilinx Heterogeneous Computing Contest is divided into registration, training, design, Q&A and finals. The finals will be held from now until April next year. The game will be based on the POWER processor and CAPI technology and the Xilinx FPGA hardware platform. The competition topics are divided into two parts: the specified proposition and the open proposition. Open proposition topics can be freely chosen in areas such as big data analysis, networking, machine learning, genetic computing, image encoding and decoding. The contestants uploaded the design plan to the SuperVessel Super Cloud (http://www.ptopenlab.com) platform created by IBM China Research Institute for verification test, and the final panel of the winners was judged by the expert review team.
Professor Wang Yu of Tsinghua University, the expert reviewer of the first IBM-Xilinx Heterogeneous Computing Contest, said: "The emergence of new workloads in the cloud computing and big data era and the limitations of processor performance and power consumption, enabling general-purpose processor computing and specific purposes. Hybrid heterogeneous computing combined with processor computing (application accelerators such as FPGA devices) is more common. Hybrid heterogeneous computing requires a combination of processor-based software programming and FPGA-based hardware programming, which brings new to traditional application programming. The challenge of the IBM-Xilinx Heterogeneous Computing Competition will bring together high-quality technical resources such as universities and open source hardware communities, which will play a positive role in the training of new programming talents. It is also of great significance for popularizing and promoting advanced second-generation distributed computing models. ."
The OpenHW open source hardware community will be the official website for this competition. For more information and registration, please visit http://www.openhw.org/events/capi.html.
CAPI technology flash open ecology casts strength
The second generation of distributed computing models will be widely used in high performance computing, cognitive computing, machine learning, genetic computing and big data analysis. With application accelerators, the throughput and latency requirements of new applications can be met at lower power. However, for application accelerators to work better with processors, there is a need to address the synergy and hybrid programming challenges of accelerators and processors.
Utilizing the advanced CAPI (Coherent Accelerator Processor Interface) technology of OpenPOWER and POWER processor, FPGA can realize efficient connection with processor, so as to make full use of the powerful hardware acceleration capability of FPGA programmable hardware to achieve soft and hard. Combine heterogeneous computing to achieve breakthrough application performance improvements.
IBM and OpenPOWER Foundation members jointly developed an application accelerator-based solution that laid the future direction of the technology industry and built a healthy and sustainable industrial ecosystem. Through an open collaborative collaboration model, OpenPOWER processors are tightly integrated with accelerators, and more than 90 members of the OpenPOWER Foundation Accelerator Working Group are developing and delivering accelerator-based solutions.
On September 16, the OpenPOWER Foundation held the second-generation distributed computing technology summit in Beijing, and established the second-generation distributed computing alliance, which was composed of processor technology vendors, server vendors, application accelerator vendors, ISVs, and industry users. The ecological chain promotes the development of technologies and applications for next-generation computing models that are characterized by the application of accelerator technology. At the same time, the CAPI FPGA App Store, which is built on the SuperVessel cloud platform, was officially launched, enabling ISVs and industry users to communicate mature application accelerator technologies and solutions more efficiently, reducing the cost and time of user adoption. Established a communication bridge to further promote the popularization and application of a new generation of computing technology in the industry.
IBM SuperVessel Ultra Cloud is a new cloud platform based on POWER technology (including Power servers and OpenPOWER servers) created by IBM China Research Institute. The platform can be used as a virtual R&D engine to learn, develop and test software applications in emerging fields such as scientific computing, big data analytics, and the Internet of Things. SuperVessel Cloud can not only provide traditional virtual computing services, but also introduce Accelerator as Service in the industry, which can provide FPGA/GPU as computing resources in the cloud to provide users with, for example, machine learning and deep learning. Service support in the fields of image, voice, and video analysis based on accelerated technology.
Jiang Xiyu, general manager of IBM Greater China Technology Strategy Cooperation, said, "The OpenPOWER Foundation is a global open research and development member organization dedicated to the open innovation of the enterprise server ecosystem, and continues to lead the development trend of computing technology. Second generation distribution Computing is becoming an irreversible trend in the industry. As one of the founding members of the Foundation, IBM is working with industry-leading technology companies and ISVs to provide the best industry for users. Apply solutions to meet the new era of computing."
Join forces to jointly promote the accelerator industry application
As a member of the OpenPOWER Foundation and the Second Generation Distributed Computing Alliance, IBM and Xilinx have worked extensively on FPGA accelerator applications. On November 16, IBM and Xilinx announced a multi-year strategic partnership that will jointly develop accelerated infrastructure, software and middleware to support emerging applications such as machine learning and network function virtualization ( NFV), high performance computing and big data analytics. On the same day, Xilinx upgraded its OpenPOWER Foundation qualification to Platinum to gain deeper involvement in the design and development of application accelerator solutions based on open data center infrastructure.
One of the most important goals of the IBM-Xilinx Heterogeneous Computing Contest is to familiarize developers who have focused on software programming and hardware programming to accommodate mixed programming environments. At the same time, the contestants will receive extensive technical support and training from IBM and Xilinx, including support from the POWER technical team and the Xilinx local Vivado SDA/HLS R&D team.
Yang Fei, vice president of Xilinx's global sales and marketing Asia Pacific region, said: "With the rapid growth of new services and massive data, the industry is increasingly demanding server performance and power consumption. Based on Xilinx fully programmable FPGA, POWER processor The second generation of distributed computing with CAPI heterogeneous computing technology is an important technology to support the development of the industry. It can bring dozens or even hundreds of times of performance-to-power ratio improvement for data center applications, while supporting the development of software-defined data centers. The world's largest fully programmable FPGA supplier, Xilinx has the industry's most cost- and power-efficient FPGA products and solutions to meet the computing performance and energy efficiency requirements of big data and new Internet applications. We believe that IBM- The Xilinx Heterogeneous Computing Contest will drive developers to become familiar with and understand new hybrid programming models faster, accelerating the adoption of advanced computing models in the industry."
