Hi , Fernando above suggestion for integrating mcpat into gem5 was useful and we were working for the patch for that. But we are just struck in one issue is that we are are not able to access the cache stats. To call mcpat dynamically we pass the stats mcpat needs after certain amount of retired instruction
So cpu stats we could access dynamically and feed into mcpat as shown below: BaseCPU *cpu = tc->getCpuPtr(); FullO3CPU<O3CPUImpl> *cpu_o3 = dynamic_cast<FullO3CPU<O3CPUImpl> * >(cpu); once we have the cpu_o3 pointer we then access stats like one shown below dynamically during run time: cpu_o3->rob.robReads.value(); Can anyone provide us how to access the cache stats. We have no clue how to get pointer to access miss,hit stats for l1,l2 for mcpat dynamically during run time. Sudarshan Graduate student Electrical Engineering Univ of Massachusetts Amherst On Fri, Nov 29, 2013 at 11:38 AM, Sudarshan <[email protected]> wrote: > > Sudarshan > Graduate student > Electrical Engineering > Univ of Massachusetts Amherst > > > ---------- Forwarded message ---------- > From: Fernando Endo <[email protected]> > Date: Thu, Nov 28, 2013 at 12:43 PM > Subject: Re: [gem5-users] Integrating gem5 with Mcpat > To: gem5 users mailing list <[email protected]> > > > Thank you! This is very usefull! > > -- > Fernando A. Endo, PhD student and researcher > > Université de Grenoble, UJF > France > > > > 2013/11/26 Andreas Hansson <[email protected]> > >> You could always give the “caching” patch from the sniper team a spin: >> http://snipersim.org/w/McPAT >> >> Andreas >> >> From: Fernando Endo <[email protected]> >> Reply-To: gem5 users mailing list <[email protected]> >> Date: Tuesday, 26 November 2013 10:47 >> To: gem5 users mailing list <[email protected]> >> Subject: Re: [gem5-users] Integrating gem5 with Mcpat >> >> Hello, >> >> It seems that one can divide mcpat execution in two or three parts: >> >> 1) chip modeling => estimates the energy cost per access >> 2) multiply accesses by energy costs >> 3) estimate the peak power of components >> >> So if you want to run mcpat faster, the step 1 can be done only once >> per gem5 simulation, this step seems to be the slower one. >> >> If you find a way to separate these, for example dumping the internal >> chip model in a file and read every time, could you please send us a patch? >> >> Regards, >> >> -- >> Fernando A. Endo, PhD student and researcher >> >> Université de Grenoble, UJF >> France >> >> >> >> 2013/11/26 Sudarshan <[email protected]> >> >>> Thanks Lluis..Only problem with this is that i need to dump stats at >>> very low instruction grnaulairty like every 5000 retired instruction and >>> the above approach makes the stats file huge(since i dump stats every 5000 >>> committed). I have already tried the above approach and its too slow. >>> >>> Thanks >>> >>> Sudarshan >>> Graduate student >>> Electrical Engineering >>> Univ of Massachusetts Amherst >>> >>> >>> On Tue, Nov 26, 2013 at 10:55 AM, Lluís Vilanova <[email protected]>wrote: >>> >>>> Sudarshan writes: >>>> >>>> > Hi, >>>> > I would like to integrate gem5 into McPAT since i would like to >>>> monitor the >>>> > dynamic power dissipated at a particular instruction >>>> granularity(example: for >>>> > every 5000 committed instructions). So i would need to call McPAT >>>> from gem5 >>>> > continuously throughout the program execution. Any pointers on this >>>> would be >>>> > helpful.I would also like to know if there is patch already available >>>> for this. >>>> >>>> The simplest way is to dump the gem5 statistics every N cycles (e.g., >>>> using >>>> 'dumpresetstats' [1]), and then process those to get numbers from mcpat. >>>> >>>> [1] http://www.m5sim.org/M5ops >>>> >>>> >>>> Lluis >>>> >>>> -- >>>> "And it's much the same thing with knowledge, for whenever you learn >>>> something new, the whole world becomes that much richer." >>>> -- The Princess of Pure Reason, as told by Norton Juster in The Phantom >>>> Tollbooth >>>> _______________________________________________ >>>> gem5-users mailing list >>>> [email protected] >>>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >>>> >>> >>> >>> _______________________________________________ >>> gem5-users mailing list >>> [email protected] >>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >>> >> >> >> -- IMPORTANT NOTICE: The contents of this email and any attachments are >> confidential and may also be privileged. If you are not the intended >> recipient, please notify the sender immediately and do not disclose the >> contents to any other person, use it for any purpose, or store or copy the >> information in any medium. Thank you. >> >> ARM Limited, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ, >> Registered in England & Wales, Company No: 2557590 >> ARM Holdings plc, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ, >> Registered in England & Wales, Company No: 2548782 >> >> _______________________________________________ >> gem5-users mailing list >> [email protected] >> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >> > > > _______________________________________________ > gem5-users mailing list > [email protected] > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users > >
_______________________________________________ gem5-users mailing list [email protected] http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
