Hello, I am getting a simliar error as this person below but I am not starting from a check point but rather just starting at the beginning of the simulation. I get the following error. The input is also show below. I am not sure of the memory size of the machine I am running it on. Could this could be what's preventing the running of the program. What else could be the issue that is causing this MMAP to return a fatal error?
Thanks, Aaron [EMAIL PROTECTED] m5-2.0b3]$ ./build/ALPHA_SE/m5.opt ./configs/example/se.py -c tests/test-progs/hello/bin/alpha/linux/hello M5 Simulator System Copyright (c) 2001-2006 The Regents of The University of Michigan All Rights Reserved M5 compiled Oct 27 2007 16:52:57 M5 started Thu Nov 1 13:41:56 2007 M5 executing on localhost.localdomain command line: ./build/ALPHA_SE/m5.opt ./configs/example/se.py -c tests/test-progs/hello/bin/alpha/linux/hello Global frequency set at 1000000000000 ticks per second mmap: Cannot allocate memory fatal: Could not mmap! @ cycle 0 [PhysicalMemory:build/ALPHA_SE/mem/physical.cc, line 66] Memory Usage: 15060 KBytes >The problem is the amount of virtual address space available to M5. >On a 32bit system that is only about 3GB. Since the simulated memory >is mapped as a contiguous block you need 2GB of contiguous free >address space. >When starting from a checkpoint the memory in the config file is >mmaped, and then when the checkpoint in unserialized, that memory is >unmmaped, the gziped memory image is uncompressed and loaded. The >problem is probably that the munmap() is before the gzdopen() so >gzdopen is probably getting a small chunk of the 2GB space which >means that you can't mmap it again. Moving the munmap() to after the >gzdopen() will probably fix the problem for you. >The other two solutions for being able to have a bigger amount of >memory are: >1. Use a 64bit system. >2. Change physical memory to allocate pages instead of the entire >memory at once and read/write them to disk as they are needed (kind >of like implementing VM in M5 for the guest OS). >Ali On Jul 26, 2006, at 11:32 PM, Adam Kaplan wrote: > I noted that on my system I am able to start the full-system > simulator and > boot Linux successfully with 2047MB of simulated DRAM (not with 2048 > though...Linux will not load), and yet when I checkpoint and try to > start > simulation from checkpoint (also with 2047MB of DRAM), I get the > following: > > fatal: Could not mmap physical memory! > @ cycle 607401203386 > [unserialize:m5/mem/functional/physical.cc, line 335] > > Anyone know a way around this? (It seems that the problem lies in > the amount of mem-alloc attempted on my system. Is the allocation > requirement different if starting from a checkpoint?) > > Thanks for any help/suggestions you can provide! > > -Adam > > ---------------------------------------------------------------------- > --- > Take Surveys. Earn Cash. Influence the Future of IT > Join SourceForge.net's Techsay panel and you'll get the chance to > share your > opinions on IT & business topics through brief surveys -- and earn > cash > http://www.techsay.com/default.php? > page=join.php&p=sourceforge&CID=DEVDEV > _______________________________________________ > m5sim-users mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/m5sim-users > ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys -- and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ m5sim-users mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/m5sim-users _______________________________________________ m5-users mailing list m5-users@m5sim.org http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
