Project Description

This project advances reduced-order, GPU-accelerated thermal simulation capabilities into practical thermal-aware applications for GPUs and AI chips. Building on an open-source, high-performance simulation framework, the work emphasizes MPI/CUDA-based scalable computation and software development in C/C++/Python (e.g., PETSc/SLEPc) to integrate simulation codes into open-source platforms and apply them to chip-level thermal management, architecture exploration, and real-time scheduling, with relevance to chip design and data-center-scale scenarios.