Brendan, Thanks for your quick reply. I really appreciate it. After receiving your response, I've tried to use -fast-fwd-checkpoint and found "checkpoint_after.c" on the internet. https://gist.github.com/avadhpatel/2218676 The source code uses -fast-fwd-chckpoint which is exactly what I need.Therefore, I compiled the source code to binary and then use ./checkpoint_after 1000M bzip2_chk ./stop_sim to create the checkpoint after 1000M user-level instructions. Then, I use qemu/qemu-systemx86-64 -snapshot -loadvm 1 -simconfig simconfig -hda CPU2006.qcow2 to boot up the VM and run the simulation.
the simconfig file is now as follows: -machine single_core -stopinsns 1000m -run Does this flow correct to get the representative result of the benchmark? Moreover, I am confused about the difference between -fast-fwd-user-insns and -fast-fwd-insns. What's the difference between user-level and kernel-level instructions? Since I'm still unfamiliar with SPEC CPU2006 benchmark, I also have some questions about it. Hope you could also help me with this problem. I run the SPEC CPU2006 benchmark by using the runspec tool provided by SPEC CPU2006 and then run the simulation as follows: ./start_sim; runspec --config=gcc_linux64.cfg --size=test --tune=base --noreportable --iterations=1 bzip2; ./stop_sim; ./kill_sim; The information I found on the internet let me know that runspec tool actually run multiple programs under specific benchmark. http://boegel.kejo.be/ELIS/spec_cpu2006/spec_cpu2006_command_lines.html Just like what the above website says, runspec tool will run bzip2 benchmark with 6 different inputs and programs. My question is whether the way I run the benchmark correct or not. Or do I only need to run one of the programs/inputs specified on the above website? Sorry for this long question description again. I have read your tutorial on the website and also googled for many times and still cannot find the answer. Any suggestion will greatly help me. Thanks a lot! Regards, Amy 2015-09-09 20:25 GMT+08:00 Brendan Fitzgerald <[email protected]>: > There's a simconfig option -fast-fwd-checkpoint that should accomplish > what you're looking to do. > > The way things work, because you put stop_insns at 200m and fast forward > at 250m, the stop gets hit first. > > You would want to do fast-forward 250m and stop_insns 450m > > On Tue, Sep 8, 2015 at 9:41 PM, 陳芷涵 <[email protected]> wrote: > >> Hello, >> I'm new to MARSS. I want to know if I can enter the simulation mode after >> the caches pass cold start phase to prevent the initialization phase from >> affecting the results? At first, I use -stopinsns 300m to stop simulation >> after 300M instructions. However, I found that 300M instructions may not be >> enough. Is -fast-fwd-insns one way to solve this problem? >> >> I've googled and tried many times about fast-fowd-insns but it still >> doesn't work. First, I boot up the VM and use ./create_checkpoint to create >> the checkpoint. >> ./create_checkpoint bzip2_chk ; <benchmark_exe>; ./stop_sim; >> Then, I create simconfig file which is shown as follows: >> >> -machine single_core >> -fast-forward-insns 250m >> -stopinsns 200m >> >> The simulation does not stop at the point it should be stopped. Following >> messages are printed again and again in ptlsim.log >> >> Stopped after 313863665 cycles, 200000001 instructions and 1134 seconds >> of sim time (cycle/sec: 276775 Hz, insns/sec: 176366, insns/cyc: >> 0.6372193512747008) >> Total Tags: single_core,ic5.theda.cad,2015-09-09,total >> Stats Summary: >> user.base_machine.ooo_0_0.thread0.commit.ipc = 1.05659 >> kernel.base_machine.ooo_0_0.thread0.commit.ipc = 0.147432 >> total.base_machine.ooo_0_0.thread0.commit.ipc = 0.637219 >> Stopping simulation loop at specified limits (313863666 cycles, 200000001 >> commits) >> >> Could you please tell me in which step I am wrong or what could cause >> this kind of problem. >> Moreover, how can I know if the fast-forwarding is successful or not. >> >> Thanks for reading this long problem description. >> I would be very appreciate if you could help me with this problem. >> Thank you very much! >> >> Regards, >> Amy >> >> >> >> _______________________________________________ >> http://www.marss86.org >> Marss86-Devel mailing list >> [email protected] >> https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel >> >> >
_______________________________________________ http://www.marss86.org Marss86-Devel mailing list [email protected] https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
