Hi Fulya,
> I am running Parsec on X86 FS mode using the precompiled Parsec binaries > for X86. When i examine the stats.txt file, I see 4 sets of simulation > statistics. I am using the .rcS script: > #!/bin/sh > # File to run the canneal benchmark > cd /parsec/install/bin > /sbin/m5 switchcpu > */sbin/m5 dumpstats > /sbin/m5 resetstats* > ./canneal 4 15000 2000 /parsec/install/inputs/canneal/200000.nets 64 > echo "Done :D" > /sbin/m5 exit > /sbin/m5 exit > I am assuming the first set of stats correspond to the m5 command > "dumpstats" that i have in .rcS script right before the benchmark starts > execution. I am also assuming that the Parsec binaries themselves have > been instrumented to dump stats based on specific regions of the program > (like ROI). Could you confirm that? Also could you tell me what the next 3 > sets of stats correspond to? I am trying to figure out how many > instructions are executed until the beginning of ROI. > That is correct that the PARSEC benchmarks on both the ALPHA and x86 disk images from our website (http://www.cs.utexas.edu/~parsec_m5/) will dump and reset statistics at the beginning and end of the benchmark region of interest. Based on your run script above, you should see 4 separate sets of statistics at the end of simulation corresponding to: (1) the start of simulation until the dump in the runscript (above), (2) the beginning of the benchmark up to the beginning of the ROI in the benchmark, (3) the benchmark ROI, and (4) from the end of the ROI to when the simulation exits on '/sbin/m5 exit' (above). If you are interested in the number of instructions leading up to the ROI, you'll be interested in the stats from the first two stats segments. Hope this helps, Joel -- Joel Hestness PhD Student, Computer Architecture Dept. of Computer Science, University of Wisconsin - Madison http://pages.cs.wisc.edu/~hestness/
_______________________________________________ gem5-users mailing list gem5-users@gem5.org http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
