In-depth understanding of both hardware and system software layers turns out to be essential. A variety of conceptual as well as technical challenges need to be addressed. Emerging compute infrastructure imposes constraints on the design of discrete algorithms. Future developments are likely to imply even more fundamental adaptations of existing and emerging algorithms. Close collaboration of developers of algorithms with their colleagues from the high-performance computer industry turns out to be crucial for the formulation of meaningful requirements on both sides. We intend to focus on sparse linear algebra and graph algorithms while reaching out to a diverse set of representatives from industry combining expertise in modern accelerators, next-generation silicon, quantum and neuromorphic computing.