Before Xilinx® Vitis™, software engineers and AI scientists working on advanced applications that demand processing performance beyond the current limits of standard CPU/GPU architectures have faced two choices to realise their vision. Either wait for the next new and most powerful compute platform to enter the market, or team-up with specialist FPGA designers to create a suitable engine using programmable hardware. Neither is ideal, imposing long delays or significant extra project expenses on innovators working at the forefront of fields such as AI inference, high-speed video encoding, genome analysis and big data analytics.
By creating the Vitis unified software platform, the power of hardware accelerators – previously available only to developers with hardware design expertise – is now accessible to a broad range of developers including software engineers and AI scientists. This enables them to accelerate their applications and adapt to rapidly evolving algorithms, industry standards and AI models at the edge, on-premise or in the cloud.
Innovation – Use Within Familiar Programming Languages
There is no proprietary development environment. Users can configure optimised compute engines in programmable hardware using familiar tools and can take advantage of a rich set of optimised open source libraries to focus on their algorithms.
Vitis enables software developers to use programming languages they already know like C and C++ to develop accelerated applications for Xilinx platforms. They can work at an application level and focus their core competencies on solving challenging problems in domain, accelerate time to insight, and innovate.
Whether they want to accelerate portions of an existing x86 host application code or want to develop accelerators for deployment on Xilinx embedded platforms, Vitis offers a design methodology that is familiar to the software developer community, without the need to learn low-level hardware implementation specifics.
Innovation – Performance-Optimised Libraries for Diverse Workloads
Vitis includes a rich set of performance-optimised libraries that can be used to accelerate a wide range of diverse applications that include AI inferencing, Quantitative Finance, Computer Vision and Data Analytics. There are more than 400 optimised functions across multiple libraries, that are completely open-source and provide different levels of abstraction and flexibility. Developers can achieve out-of-box acceleration with minimal to zero-code changes to their existing applications, written in C, C++ or Python or modify these libraries to suit their specific applications’ needs.
Innovation – Domain-Specific Development Environments
Xilinx is further extending Vitis through domain-specific development environments that connect with industry-leading frameworks for machine learning, video encoding, and analytics. The first of these is Vitis AI, which integrates with industry-leading frameworks like TensorFlow and Caffe, for AI inferencing on Xilinx platforms. Vitis AI provides the tools to optimise, compress and compile trained AI models running on a Xilinx device in about one minute. It also delivers specialised APIs for deployment from edge to cloud, all with best-in-class inference performance and efficiency.
Xilinx is also partnering with industry-leading specialists to make additional DSAs available. These include Illumina, which integrates with GATK for genome analysis, and BlackLynx, which integrates with ElasticSearch for big data analytics.
Innovation – Comprehensive Developer Tools
The Vitis core development kit contains tools that are designed to integrate seamlessly with industry-standard build systems and development environments. These include the open-source Xilinx runtime library to manage the data movement between different domains, including the subsystems, the AI Engine in the Versal ACAP™, as well as an external host, if required. This layer also includes a complete set of graphical and command-line developer tools comprised of the Vitis compilers, analysers and debuggers to build, analyse performance bottlenecks and debug accelerated algorithms. Users can leverage these features within their own IDEs or use the standalone Vitis IDE.
The base layer is the Vitis target platform, which defines base hardware and software architecture and application context for Xilinx platforms, including external memory interfaces, custom input/output interfaces and software runtime.
Faster Solutions to Critical Challenges
By allowing a familiar development environment, providing open-source libraries that simplify access to ready-to-use applications, and allowing future extensibility to keep pace with the forefront of high-performance software and AI development, Xilinx Vitis removes a notorious brake on technological progress and provides a platform for developing solutions to today’s urgent challenges in healthcare, energy, industry, cyber-security, transportation, and smart living.