Project Grant 2551122

The National Science Foundation Division of Information and Intelligent Systems awarded the University of California, Los Angeles $500,000 on October 1, 2026, under the Computer and Information Science and Engineering program (CFDA 47.070) to develop a complete system stack for automated compute scaling using Compute Express Link (CXL) shared memory.

The project delivers four technical tasks. The first establishes fine-grained sharing of individual memory locations with safe synchronization abstractions. The second creates compiler and runtime support that dynamically migrates program objects across different memory regions. The third implements transparent page-based memory sharing inside the operating system to enable unmodified applications to scale automatically. The fourth builds a dynamic analysis tool that detects coherence errors in software relying on software-coherent memory.

The research addresses the challenge that cloud computing programs typically built for single machines must be extensively rewritten to run across multiple servers. By enabling existing programs to execute across many machines without modification through CXL-based shared memory with cache coherence, the project aims to improve server utilization, reduce hardware overprovisioning, and lower data center operating costs. The benefits extend particularly to research laboratories, academic institutions, and small companies lacking large infrastructure but requiring scalable computing capabilities.

Performance occurs at UCLA's Los Angeles campus (90024) with a period of performance from October 1, 2026, through September 30, 2030.