Are you running with the -snapshot option? If so, that is why the
filesystem isn't changing after the run. You have to explicitly do a
commit in the QEMU monitor to make the changes persistent.

Also, keep in mind that if you are writing to the filesystem in the
simulated guest machine, you are going to create a significant amount of
noise in the results since that requires the guest OS to do all kinds of
work. If you simply want to be able to see the output, take a look at the
ptlcall_log() function (implemented in ptlsim/tools/ptlcalls.h and
processed by marss in ptlsim/sim/ptl-qemu.cpp). It is very low overhead
from the simulated machine's perspective because it simply passes a
pointer down to PTLSim and then the simulator grabs the data in one
simulated cycle and prints it out to the marss log file. If you label the
data you are dumping out to the marss log file, it is trivial to extract
it via post processing. Or you could add a ptlcall to dump it to a
particular output file. You could then write a small C program that dumps
everything it receives on stdin to ptlcall_log() and back to stdout and
then just pipe the output of your benchmark into the logger program. This
will let you see the output in the guest while simultaneously dumping it
to the host.

Hope that helps.

Jim Stevens
University of Maryland, College Park

> Hi,
>
> How can I write(keep) the output of a benchmark into a txt file when I run
> it in the MARSS? I tried the following command chains in both .raw and
> .qcow2 disk images, however, after rebooting the images, neither of them
> contains any ls1.txt file.
>
> Thanks
>
> Command Chains:
>
> ./start_sim; ls > ls1.txt; ./kill_sim
>
> ./start_sim; ls > ls1.txt; ./stop_sim; ./kill_sim
> _______________________________________________
> 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

Reply via email to