CuClarabel - GPU Acceleration for a Conic Optimization Solver

Learn about the GPU implementation of Clarabel, a general-purpose interior-point solver for convex optimization problems with conic constraints, in this 35-minute conference talk. Discover how a mixed parallel computing strategy processes linear constraints first before handling other conic constraints in parallel, with current support for linear, second-order cone, exponential cone, and power cone constraints. Explore how integrating this mixed parallel computing approach with GPU-based direct linear system solvers significantly enhances performance compared to CPU-based counterparts across diverse conic optimization problems. Understand the potential for additional acceleration through mixed-precision linear system solvers while maintaining solution accuracy, providing valuable insights for researchers and practitioners working with large-scale optimization problems in Julia.