Hi all, I'm running a mix of two benchmarks from PARSEC on FS and Ruby memory system, here is my script:
#!/bin/sh # File to run the dedup benchmark cd /parsec/install/bin /sbin/m5 switchcpu /sbin/m5 dumpstats /sbin/m5 resetstats taskset 0X1 ./dedup -c -p -f -t 1 -i /parsec/install/inputs/dedup/medias.dat -o /parsec/install/inputs/dedup/output.dat.ddp & taskset 0X2 ./canneal 8 10000 2000 /parsec/install/inputs/canneal/100000.nets 32 echo "Done :D" /sbin/m5 exit /sbin/m5 exit And here is my command: ./build/X86/gem5.opt ./configs/example/fs.py --cpu-type=timing -n 2 --ruby --script=/path/to/script In the stat.txt file, I have 5 sets of results, I think that third and fourth part of results should be related to ROI parts of two benchmarks, am I correct? I've changed cache coherency protocol and now I want to see it's effect on performance (IPC, for example). I know that IPC is not reported in timing cores simulation, but is there any way to calculate it manually? I calculated IPC based on the value of committed instructions and number of cycles, but it's low (around 0.2)! I don't know what is wrong? I know that in timing mode, each instruction simulates in one cycle, and memory requests take more than one cycle, but why is it too low? I tried to use detailed mode, also. But as I posted before ( http://www.mail-archive.com/gem5-users@gem5.org/msg13126.html), it was not successful. Any help in calculating IPC is appreciated. Thanks in advance, Hodjat
_______________________________________________ gem5-users mailing list gem5-users@gem5.org http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
