Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only!
Explore a cutting-edge compiler-directed methodology for optimizing performance in multi-core embedded systems through this 22-minute video presentation from the LCTES 2023 conference. Dive into the innovative PinIt approach, which analyzes applications individually to induce operating system-mediated actions across applications, minimizing harmful migrations and maintaining data locality. Learn how PinIt determines high memory reuse density regions and strategically places pin/unpin calls to avoid excessive cache and TLB misses. Discover the new analyses and transformations that optimize call placement, and examine the impressive performance improvements achieved in Mediabench and vision-based workloads compared to priority-cfs. Gain insights into addressing the challenges of thread migration in complex co-execution environments, particularly for embedded servers running media and vision applications with high data locality requirements.