Improving the Life-Cycle of Tensor Algorithm Development with ITensors

Learn about advancing tensor algorithm development in this JuliaCon 2024 conference talk that addresses the challenges and solutions in creating robust software frameworks for modern computational resources. Explore how the ITensors software suite tackles complex issues in algorithmic development, from memory-distributed algorithms to processor-specific language differences. Dive into the implementation of the NDTensors module, which employs generic programming practices to handle dense and sparse tensor arithmetic, buffered memory management, linear and multilinear algebra, GPU implementations, and task management. Discover practical applications through demonstrations of accelerating DMRG optimization for Hubbard and Heisenberg models, showcasing how minimal code modifications can leverage GPU accelerators effectively.