Re: [perfmon2] Implementing Intel's drill-down technique for performance analysis
Status: Beta
Brought to you by:
seranian
From: stephane e. <er...@go...> - 2010-11-25 15:44:11
|
Hi, What machine are you trying to do this on? On Thu, Nov 25, 2010 at 4:39 PM, Eduard Diner <edu...@go...> wrote: > Hi guys, > I have instrumented my code on Nehalem with libpfm4 and now I am a little > bit confused. > I am reading Intel's "IA64-Optimization-Reference". In Appendix B3.4 they > describe a drill-down technique for performance analysis. I think this is a > good starting point for analysis. > Basically I want to map Intel's description in the document to the events of > libpfm4. This sounds easy... but it is not that easy for me :-) > > > Intel writes for example: > "Cycles_non_retiring_uops .... A constant issue rate of μops flowing through > the issue port. Thus, we define: > uops_rate = Dispatch_uops/Cycles_issuing_uops, where Dispatch_uops can be > measured with RS_UOPS_DISPATCHED, clearing the INV bit and the CMASK." > > Well, I don't see a RS_UOPS_DISPATCHED event in libpfm. As far as I > understood the architecture, UOPS_DISPATCHED can be expressed as the sum of > UOPS_RETIRED and UOPS_ISSUED. > What do you think? > > Regards, Eduard > > ------------------------------------------------------------------------------ > Increase Visibility of Your 3D Game App & Earn a Chance To Win $500! > Tap into the largest installed PC base & get more eyes on your game by > optimizing for Intel(R) Graphics Technology. Get started today with the > Intel(R) Software Partner Program. Five $500 cash prizes are up for grabs. > http://p.sf.net/sfu/intelisp-dev2dev > _______________________________________________ > perfmon2-devel mailing list > per...@li... > https://lists.sourceforge.net/lists/listinfo/perfmon2-devel > > |