Just to clarify, running scripts in SE in the same way that the .rcS scripts are run under FS (i.e., inside the simulated system) only makes sense if your goal is to use the python interpreter as a benchmark. Running in SE mode basically means you're running what would be a single process in a real system. If you want script-based control over the process that gets run (which is what the .rcS file is for) then you need to move that scripting code into your simulation control script (e.g., se.py).
That said, if you want to run PARSEC, you're better off running in FS mode anyway, since the single-process multithreading and inter-thread synchronization support in SE mode is not really that great. That's why all the current work has been done in FS mode. Steve On Fri, May 21, 2010 at 10:29 AM, Gabe Black <[email protected]> wrote: > The cvti2f uop was fixed a little while ago so you won't need that > patch. FS mode is set up to run scripts inside a simulated OS > environment. In SE mode your running a program directly so that doesn't > quite make sense. You could run the script interpreter as the binary and > then feed it the script as input, although it would have to be > statically linked. To the best of my knowledge all the ISAs we support > are supported in SE mode since that's generally where the work starts, > and they all work with se.py. > > Gabe > > Vince Weaver wrote: >> On Fri, 21 May 2010, Malek Musleh wrote: >> >>> In the long run, i am interested in running the parsec benchmarks on >>> the X86 ISA, (SE simulation is fine) and would like to know if the >>> current stable version requires relatively modest >>> modifications/improvements for my studies, and if not, if the current >>> dev-repo is a viable alternative. >>> >> >> you're going to need to run the development tree. >> >> You'll also need additional patches. You can see here: >> http://www.csl.cornell.edu/~vince/projects/m5/m5_x86_64_se_status.html >> for the status of m5 for x86_64, at least with the SPEC2000 benchmarks. >> >> There's at least 8 patches or so there that would need to be applied >> before SPEC2000 ran, I've been meaning to clean them up and submit them at >> some point, but graduating/getting a new job has sucked up all my time. >> >> In general, none of the x87 floating point support will work. You'll need >> to compile your binaries using SSE only. >> >> I also am not sure what the status of multi-threaded programs is in X86_SE >> mode, I imagine it will take a bit of work to get stock compiled PARSEC >> running. >> >> Vince >> >> _______________________________________________ >> m5-users mailing list >> [email protected] >> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users >> > > _______________________________________________ > m5-users mailing list > [email protected] > http://m5sim.org/cgi-bin/mailman/listinfo/m5-users > _______________________________________________ m5-users mailing list [email protected] http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
