Big memories on the desktop

Abstract

Operating systems people think of desktop systems as "small" machines and worry about supporting small main memories. This is a historical aberration; memory sizes (and memory demands) are growing faster than memory access times are decreasing, so desktop operating systems will have to do a better job at supporting large memories. I discuss three problem areas for programs with large address spaces: avoiding TLB misses, better use of backing store, and use of explicit control to avoid paging latencies.<>