Sorry, I haven't looked at the specific script you're using, though if it's an example script I'd hope they have set it up correctly.
Usually you have to do some work to signal to the library how many threads you want to run. For example, for OpenMP, you'd set the environment variable OMP_NUM_THREADS=4 to signal to the runtime libraries that you want 4 threads. Pthreads code takes a more explicit approach where the application figures out how many threads to spawn. On Tue, Aug 13, 2013 at 10:14 AM, Jia Zhan <[email protected]> wrote: > > > On Mon, Aug 12, 2013 at 9:29 PM, Jia Zhan <[email protected]> wrote: > >> Hi Paul, >> >> I am sorry but I don't quite get it. How do I assure the applications are >> using the extra cores? >> >> FYI, here is what I used in the command line to run blackscholes with 4 >> cores: >> >> %./build/ALPHA_FS/gem5.opt configs/example/ruby_fs.py >> --script=./PARSEC/4-core/blackscholes/blackscholes_4c_simsmall.rcS >> --cpu-type=timing --caches --l2cache --num-cpus=4 --num-dirs=4 >> --num-l2caches=4 --cpu-clock=1GHz --l1d_size=64kB --l1i_size=64kB >> --l1d_assoc=2 --l1i_assoc=2 --l2_size=4MB --l2_assoc=8 --cacheline_size=64 >> --ruby --garnet-network=flexible --topology=Mesh --mesh-rows=2 >> >> Thanks, >> >> Jia >> >> >> On Mon, Aug 12, 2013 at 9:00 PM, Paul Rosenfeld <[email protected]>wrote: >> >>> Are you certain that the applications are actually using the extra cores >>> (i.e., are you setting flags for parsecmgmt like -n)? If a single thread >>> runs on more simulated cores, I'd imagine that the runtime would be >>> identical while the number of instructions would go up (due to idle threads >>> running on the spare cores). >>> >>> >>> On Mon, Aug 12, 2013 at 3:38 PM, Jia Zhan <[email protected]> wrote: >>> >>>> Hello, >>>> >>>> I am running a single benchmark with different number of cores to >>>> observe the performance speedup. However I didn't see any reductions in >>>> *sim_second*s when the number of cores increases from 1 to 16. Also, I >>>> noticed that* sim_insts* (number of instructions simulated) also >>>> increase as the number of cores increase. Isn't true that the number of >>>> instructions should stay the same for a specific benchmark? >>>> >>>> Best, >>>> Jia >>>> >>>> _______________________________________________ >>>> 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
