Hi everyone!
   I've been trying to run as many PARSEC benchmarks as I can in sys call 
emulation mode with proper compilation and with m5threads and --static. There 
are some peculiarities in specific benchmarks that I was hoping some people in 
here have faced already :-) I couldn't really find any answers in mail archives.
I'm hoping to use sys call emulation mode because full system takes a few days 
to boot up the linux kernel... I am not sure if this is normal but this is what 
I am experiencing with gem5-dev right now.
Blacksholes, Fluidanimate, freqmine, steamcluster, x264 all run well. 
Bodytrack, Vips and ray trace have some static library dependency issues in my 
machines. However:


Canneal: Terminates with:warn: mmap length argument 0x7ffff7000 is unreasonably 
large.fatal: Out of memory, please increase size of physical memory.(this is 
similar to swaptions. Does this refer to the host OS (i.e. gem5 needs more 
memory), or does it mean I need to make the simulated DRAM larger?
Facesim. Runs but terminates with:warn: This sys_clone implementation assumes 
flags CLONE_VM|CLONE_FS|CLONE_FILES|CLONE_SIGHAND|CLONE_THREAD (0x10f00), and 
may not work correctly with given flags 0x100011panic: Tried to write unmapped 
address 0x168.The closest I could find is some old emails saying that it's 
because the host architecture is 64-bit but I can't really compile it as 32bit 
due to other library dependencies.
Ferret: Terminates with:panic: Tried to read unmapped address 0x20. @ tick 
2064240
Swaptions: fatal: Out of memory, please increase size of physical memory. @ 
tick 2090024[allocPhysPages:build/X86_MESI_Two_Level/sim/system.cc, line 
323]Memory Usage: 2090024 KBytes

Thanks in advance!  George M
_______________________________________________
gem5-users mailing list
gem5-users@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users

Reply via email to