The Linux Foundation Member Summit

The landscape of computing has changed and the majority of software developers are looking to use architectures with multiple types of processors, in particular GPUs. The emergence of modern AI techniques has further accelerated the usage of GPUs and new AI processor architectures to achieve the performance required by neural networks. The challenge developers are seeing is that different architectures often require unique languages, tools and libraries, adding complexity and limiting code reuse. The solution is to provide a single way to develop across these architectures, and the basis for this must be rooted in an open, standards-based programming model and libraries. The UXL Foundation was formed to tackle this, and we are working together to build a solution to these challenges – providing an open, standards-based way to develop software and deploy across vendors. This presentation will show how the UXL Foundation is evolving its projects to meet the vendor neutral demands of software developers by adopting open standards and open source, and how our members are using the projects to enable projects like PyTorch and TensorFlow.