Program/Compiler Optimization
Details and sub-projects for Program/Compiler Optimization.
Program/Compiler Optimization
Understanding Programs from Architectural Perspective
Research Overview
Program and compiler analysis from an architectural perspective improves optimization across heterogeneous systems. The work develops robust program representations, handles proprietary and perturbed input/architecture settings, and builds methods for representative program generation that reflect real workloads. It also investigates semantic and syntactic similarity with contrastive learning while pursuing automation for program/compiler optimization and better HPC-cloud efficiency through dynamic resource-utilization profile based scheduling.
Research Details
Current efforts in this area include Representation Learning for Architectural Program Understanding: Learn architecture-aware program representations for optimization and analysis. , Program Similarity via Contrastive Learning: Similarity definitions driven by semantic/syntactic structure. , Automatic Program/Compiler Optimization for Heterogeneous Systems: Using AI and automation to migrate between instruction sets. , and HPC and Cloud Efficiency via Dynamic Profile-Based Scheduling: Improve resource utilization and scheduling outcomes in production systems. .