c - Detecting Cache Misses and Hits Pragmatically in Linux -
I know this is a platform-specific question, however, I want to see some run-time analysis of an application Cash recall and hit I know that Cachegrind, a device of Valgrind, and vtune, and other profiling utilities have a very existent, however, I'm interested in implementing my own version of Cash-Missed Detection I know cachegrind Cash-sim Without serves as Uletr separating the kernel, I can find out how cash-Miss practical? I have a feeling that user-land is almost impossible for application, but I had to ask anyway.
How can I detect cash-miss practically [without cash simulation] Am I
Cache is managed by hardware - not the kernel, their parameters (cache, size, kick-out policy, list-back / list-through, etc.) all processor implementation-specific Are there. As a programmer, you should not "know that they exist" In this way, it is impossible to measure cash-missed performance without a cash simulation.
On the other hand VM page (one lot roughly "cache" - in the sense that one of the cache memory is managed by the OS) I imagine that page defects To collect statistics about hacking in the kernel or even creating a Nifty user application. Page fault data can not be that too much use for you (especially since they are affected by other running processes), but large amounts of RAM may be (A small, small bit) page-fault or page-access pattern similar to the CPU cache access pattern, however, I'm not sure about the details.
Comments
Post a Comment