Algorithm-Level Programmability™ Approaches Fixed Function Power, Performance and Area Efficiency
Santa Clara, CA – October 7, 2015 – Vivante® Corporation announces the immediate availability of the VIP7000 family of Vision Image Processor IP cores. The VIP7000 has been designed into a range of products from a mass market IoT surveillance client SoC to an automotive imaging application. The VIP7000 features Algorithm-Level Programmability where the basic element of application development is at the algorithmic level rather than the arithmetic level. This model provides performance and power consumption approaching fixed function with rapid adaptation for new or updated applications. Contrasted with arithmetic-level approaches, applications can be developed more quickly and consistently with less silicon area and power consumption.
“The challenges we see in the IoT camera market are the conflicting demands for small silicon area and a growing set of feature requirements. For example, the analytics requirements for cloud aggregation in a retail environment versus a parking structure are typically distinct.” said Shanghung Lin, Vivante’s Director of Vision Algorithm and Application Research and Development. “The market requires a solution that is silicon-efficient and adaptable to both varied and unanticipated use cases. Algorithm-Level Programmability is a technology that not only provides a solution to these challenges, but reduces the complexity of application development and enables Vivante to deliver advanced capabilities such as deep neural networks.”
In areas such as computer vision, where both applications and algorithms are evolving, the cost and time required to implement new silicon is increasing. This challenge is compounded by sensor and screen resolutions increasing more rapidly than on-chip resources. The VIP7000 architecture provides both configurable instructions and hardware acceleration elements to enable new functionality in existing silicon. VIP7000 provides a method to incorporate parallel computing advances in software rather than new silicon. The VIP family features a stream interface, which enables algorithm optimization across module boundaries. For example, the VIP low latency stream interface enables the insertion of a programmable vision module into the fixed function image signal processing (ISP) pipeline.
“We developed the VIP family of products to address the challenges of both mass market IoT and automotive applications that were underserved by existing technology,” said Wei-Jin Dai, president and CEO of Vivante. “Vivante made a major commitment to computer vision solutions several years ago. The enthusiasm of our lead partners has been both gratifying and a validation of our strategy and the reward for our investment in this segment.”
Vivante’s VIP family was one of the first to pass the OpenVX™ conformance test. The industry is rapidly adopting OpenVX as a standard API for vision programming. OpenVX and OpenCL™ are important API standards helping to expand the computer vision ecosystem and are a key component in the VIP product family.
About Vivante Corporation
Smaller – Faster – Cooler: Vivante Corporation, a leader in dedicated Vision Processing and multi-core GPU, Compute, CPC Composition Engine and Vector Graphics IP solutions, provides the highest performance and lowest power silicon characteristics across a range of Khronos Group API conformant standards based on the Vega architecture. Vivante GPUs are integrated into customer silicon in devices targeting wearables, IoT, smartphones, tablets, HDTVs, automotive, consumer electronics and other embedded devices and running thousands of graphics/compute applications across multiple operating systems and software platforms. Vivante is a privately held company headquartered in Santa Clara, California, with additional R&D centers in Shanghai and Chengdu.
