Messages in this thread | | | Date | Fri, 22 Apr 2011 21:57:20 +0200 | From | Ingo Molnar <> | Subject | Re: [generalized cache events] Re: [PATCH 1/1] perf tools: Add missing user space support for config1/config2 |
| |
* Andi Kleen <ak@linux.intel.com> wrote:
> > Once you have clear and precise definition, then we can look at the actual > > events and figure out a mapping. > > It's unclear this can be even done. Linux runs on a wide variety of micro > architectures, with all kinds of cache architectures. > > Micro architectures are so different. I suspect a "generic" definition would > need to be so vague as to be useless.
Not really. I gave a very specific example which solved a common and real problem, using L1-loads and L1-load-misses events.
> This in general seems to be the problem of the current cache events. > > Overall for any interesting analysis you need to go CPU specific. Abstracted > performance analysis is a contradiction in terms.
Nothing of what i did in that example was CPU or microarchitecture specific.
Really, you are making this more complex than it really is. Just check the cache profiling example i gave, it works just fine today.
Thanks,
Ingo
| |