Session

DescriptionMoore’s Law has permitted computer architects to double system performance every 18 to 24 months, enabling dense, compute-bound applications to achieve similar gains. However, an important class of data-driven applications dominated by artificial intelligence are evolving and scaling to models with trillion of parameters. These AI applications are revolutionizing and empowering applications, but are often coupled with analytic computations on graphs that are communication-bound dominated by irregular memory accesses; consequently, government and industry sponsored projects supported by academic research teams have emerged to co-design next generation computer systems that address the challenges of these disruptive applications. But too often these programs are insular, with limited collaboration and cross-fertilization opportunities. This symposium provides a forum for performers to meet and discuss their visions and ideas for the development of novel architectures, algorithms, and software methodologies to enhance the performance and efficiency of large-scale HPC systems for data-driven applications.