Hi Meng-Ju, Thanks for creating that page. It's great and will probably help a lot of people. Looking at the development repository we've implemented ftruncate(), getrlimit(), and getcwd(), but we don't yet have code for mremap(). I thought about including your code, however I don't understand what it's doing.
It seems like the simplest implementation is to do what you've done, just allocate a new chunk of memory and say the start address has moved. This leaks memory because the previous memory will never be reclaimed, but it might be ok. I don't quite see the code you have on the wiki doing this though. You allocated a new region that is new-old large and place it at the end of the mmap and then set the start address to the original address. This will work if mremap() is remapping the previous mmap(), but otherwise it won't work. Minimally, there probably needs to be a check that verifies that is what's happening (I think start + old_size == p->mmap_end will do it) and if not re-allocate the entire region. There probably should also be a panic() in there if mremap is called more than a handful of times, as it would start to create a memory footprint that didn't match reality. If you have a few minutes could you address that problem and clean the code up a little bit. I would like to commit it to the repository so that people won't have to patch m5 to get the spec cpu 2006 benchmarks to work. Thanks, Ali On Sep 29, 2008, at 3:57 PM, Meng-Ju Wu wrote: > Hi all, > > I add a page in M5 wiki FAQ to show how to run the SPEC2006 on M5 SE > mode. I am focusing on parallel programs now, so I don't have much > time to work on the SPEC2006. I re-simulated the whole SPEC2006 last > weekend to test the testing data set and get the timing results. This > wiki page shows what we have done so far at UMD. I hope our > experiences can help everyone who wants to simulate SPEC2006 on M5. > > Meng-Ju > _______________________________________________ > 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
