| 摘要 |
Fueled by higher clock rates and superscalar technologies, growth in processor speed continues to outpace improvement in memory system performance. Reflecting this trend, architects are developing increasingly complex memory hierarchies to mask the speed gap, compiler writers are adding locality enhancing transformations to better utilize complex memory hierarchies, and applications programmers are re-coding their algorithms to exploit memory systems. All of these groups need empirical data on memory behavior to guide their optimizations. This paper describes how to combine simple hardware support and sampling techniques to obtain such data without appreciably perturbing system performance. By augmenting a cache miss counter with a compare register and interrupt line such that the processor is interrupted when the counter matches the compare value, we can sample system state and develop cache miss profiles that associate cache misses with specific processes, procedures, call stacks, addresses, or user defined aspects of system state. This idea is implemented in the Mprof prototype that profiles data stall cycles, first level cache misses, and second level misses on the sun Sparc 10/41. Simple case studies are provided to illustrate Mprof's features. |
