Hello everyone, I'm trying to run different applications for ARM full-system mode using big.LITTLE_2_2 configuration with --cpu-type=exynos in order to model the Odroid-XU3 board within gem5 environment. I want to explore different performances and power consumption when the application itself is pinned just to LITTLE cores or big ones.
However two issues arise: #1) How can I recognize which cpu_id is the big or LITTLE? Running /proc/cpuinfo within the fs-mode, all the 4 cores have the same part number (CPU part) so I can't decide beforehand which core_id number I need to use for running just the big or LITTLE core. #2) Following the order used on the .dtb filename, let's suppose that the first 2 cores (i.e. cpu0 and cpu1) are the big one, whereas cpu2 and cpu3 are the LITTLEs. Running the benchmark on the last two cores, I would expect from stats.txt result a very low IPC from the two big cores as they are not involved in the execution. However, this statistical event for bigCluster assumes values between 0.4 and 0.6, comparable with littleCluster as they are involved in the execution too. The same happen for all the other relevant parameters. How is that possible? I can provide further informations about the current setup that I'm using. Any help or suggestions would be greatly appreciated. Regards, Francesco Mazzei
_______________________________________________ gem5-users mailing list gem5-users@gem5.org http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users