Microsoft officials have been talking up, as of late, the need for some fundamental changes in software design and development to accommodate the shift to many cores on client systems. There are several Microsoft Research projects focusing on various aspects of the many-core challenge. The newest and seemingly most far-reaching of these is MS-ManiC.
According to a new page on the Microsoft Research site, the MS-ManiC (Memory Systems for Many Cores) project
- “is focused on designing scalable memory system architectures for future many-core processors. The memory system is not only one of the major performance bottlenecks in chip multiprocessors, but its design can lead to serious energy inefficiency, unpredictability, and security holes. Our goal is to design a scalable and balanced memory system using a holistic hardware/software approach. We are interested in designing high-performance, energy-efficient, and secure memory system that at the same time provides quality of service to applications utilizing it.”
Sounds ambitious. What form would such a memory system take and when might it move from research to a product group? No word. In fact, there’s very little information (yet) on the MS-ManiC site, other than a link to a newly published whitepaper, “Memory Performance Attacks: Denial of Memory Service in Multi-Core Systems,” (which is part of the proceedings of the 16th USENIX Security Symposium from August 2007).
Microsoft, MS-ManiC, Microsoft Research, Microsoft Projects, Multicore