Thanks for static part knowledge. And Today I accidentally discover what was the reason behind Why I was not getting correct virtual address trace as expected. Early I had installed 64 bit ubuntu on VM player. As, today my system crashed and I re-install every thing. GET RID OF VM AND installed 64 bit ubuntu this time.
Not sure but since VM does the binary translation and VM interally does blocking of group of instruction, same as PTLsim does for faster decoding/pre-decoding of instruction. Thats why I was getting same virtual address for different x86 instruction and even the address were not having one to one correspondence with the bench mark I was running. NOTE: Avoid running marss on top of Virtual machine.!!!!!!!!!!!!!!!!!!!!!! Thanks & Regards Sunil Shah On Tue, Jan 18, 2011 at 11:39 PM, avadh patel <[email protected]> wrote: > > On Tue, Jan 18, 2011 at 3:59 AM, sunil shah <[email protected]>wrote: > >> Thanks a lot Avadh. >> >> >> On Tue, Jan 18, 2011 at 12:06 AM, avadh patel <[email protected]>wrote: >> >>> Nope, sequential core of PTLsim does not work with Marss. But we do have >>> an Atom like Core model in 'core-models' branch which is work in progress. >>> You can checkout that branch from github repository and take a look at >>> 'ptlsim/core/atom-core/' for code. >>> >>> I will look explore this Atom like core soon. >> >> >> >>> If you are only interested in getting function trace, you can enable >>> 'TRACE_RIP' in 'ptlsim.h' file and recompile to get execution trace in >>> 'ptl_trace_rip' file. This will dump the whole execution trace. Take look at >>> this script >>> https://github.com/downloads/avadhpatel/marss/trace_to_func.py for >>> mapping this execution trace to functions. >>> >>> All this I did for blackscholes parsec benchmark. >> As You suggested i did and generated the ptl_rip_trace and apart and then >> finally able to get the function call graph after running above mention >> python script. Resutl what i got is only "main" function entry and other >> time <cannot find the function>. >> Not sure why only entry for function main is there in the result and apart >> from it, I am trying to get basic block trace not function call graph trace. >> So I am looking for more fine granuled rip trace. >> >> To get rid of <cannot find the function> you will need to compile the > binaries as 'static' so all the library functions are compiled in. > For kernel level functions if you don't have object dump information of > running kernel then it will also show up as <cannot find function>. > > - Avadh > > >> Thanks & Regards >> Sunil shah >> >> >> - Avadh >>> >>> On Mon, Jan 17, 2011 at 4:11 AM, sunil shah >>> <[email protected]>wrote: >>> >>>> Hello, >>>> >>>> I am not sure and did not found any mention of running MARSS with >>>> sequential core? Did anyone had tried to run MARSS of sequential core of >>>> PTLSim. >>>> If not then is it possible to run MARSS in with sequential core of >>>> PTLSim to get the functional trace. >>>> >>>> >>>> Thanks in anticipation. >>>> >>>> Regards >>>> Sunil Shah >>>> >>>> >>>> _______________________________________________ >>>> http://www.marss86.org >>>> Marss86-Devel mailing list >>>> [email protected] >>>> https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel >>>> >>>> >>> >> >
_______________________________________________ http://www.marss86.org Marss86-Devel mailing list [email protected] https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
