- This event has passed.
OFI Libfabric, APIs for MPI & CCL
Presenter Name: Rajalaxmi Angadi, Juee Himalbhai Desai, and Peinan Zhang (Intel)
Abstract: With increase in AI application and optimizations in HPC/AI, there are multiple middleware libraries using libfabric. This tutorial will go through libfabric overview, it will explain the software stack hierarchy. There will be examples programs with point to point and collectives coming from middleware library. There will be demonstration running simple programs on CPU/GPU.
Bios: Rajalaxmi Angadi is a seasoned engineering manager and leader with over 19 years of experience in software industry. She excels in the productization of OFI libfabric and the management of cluster infrastructure for HPC and AI. Her career encompasses roles in leadership, architecture, software development and performance engineering. She has made significant contributions to industry standards such as OFA, UEC, NVMe, and DMTF SNIA, demonstrating her dedication to technological innovation and advancement of cutting-edge technologies. She co-chairs Technical Women at Intel Network (Senior Women Community & Network of Executive Women), fostering the development of women in technology.
Juee Himalbhai Desai is a Middleware Development Engineer at Intel Corporation to support libfabrics software productization. She has 4 years of industry experience working in computer networking. She has been with Intel developing and testing libfabric components for the past 3 years. Prior to Intel, she was with Pluribus Networks working on physical layer technologies.
Peinan Zhang is a senior software developer at Intel with over 20 years of experience in developing performance tools and runtime libraries. Her current focus is on high-performance computing networking solutions and performance analysis technologies, providing Open Fabric Interface (OFI) and profiling tools to assist developers in building robust and scalable HPC and AI applications. Peinan obtained a PhD in Computer Science from Purdue University.