Are the cores that have 0 instructions assigned to gobmk and h264ref? I think what may be happening is that you're assigning the same instance of the benchmark object to multiple CPUs, which is the right thing to do for a multi-threaded benchmark, but for a single-threaded benchmark that means that all the CPUs but one will never get started up. You probably need to add parenthesis [Mybench.h264ref()] to create a new instance of the object for each CPU.
That still doesn't explain why it does work correctly in one case, but it's consistent with the problems you're having. Steve On Tue, Mar 3, 2009 at 11:06 AM, Steve Reinhardt <[email protected]> wrote: > I mean output from the benchmarks, not output from the simulator. > Where this goes is controlled by the 'output' parameter on the Process > objects. Look in your config.ini if you don't know what this is set > to. > > Also, using 60GB of physical memory for your simulation is somewhat > suspicious... unless you have that much physical memory or swap on the > actual machine you are running the simulations on I wouldn't expect > that to work. > > Steve > - Show quoted text - > On Tue, Mar 3, 2009 at 10:04 AM, Hari Kiran <[email protected]> wrote: >> >> hi!!! >> >> I don't find any error messages in the output. I am using 60GB of physical >> memory which is more than sufficient I guess. >> The output can be seen in the attached files for two cases. >> >> >> On Wed, Feb 25, 2009 at 9:07 PM, Steve Reinhardt <[email protected]> wrote: >>> >>> Have you looked at the program outputs to see if there are any error >>> messages? I have an inkling that if you're trying to run a large >>> number of SPEC 2006 benchmarks concurrently you may be running out of >>> (simulated) memory. That doesn't explain why it works with one >>> ordering and not with another, but it's my best hypothesis without any >>> more data. >>> >>> Steve >>> >>> On Wed, Feb 25, 2009 at 6:40 PM, Hari Kiran <[email protected]> wrote: >>> > >>> > >>> > hi! >>> > >>> > I am running eight-core system simulation for integer benchmarks. I have >>> > two >>> > cases mentioned below . In case(a) i have assigned different benchmarks >>> > for >>> > different cpu's and in case (b) I changed the order of assignments of >>> > benchmarks for different cpu's. I got the results for case(a) but in >>> > case(b) >>> > I got some cpu's values to be 0 in m5stats.txt file. What might be the >>> > problem ? >>> > >>> > case (a) : >>> > >>> > max_insts_all_threads = 100000000 >>> > num of cpus = 8 >>> > num threads per cpu = 1 >>> > system.cpu[0].workload = [Mybench.perlbench] >>> > system.cpu[1].workload = [Mybench.h264ref] >>> > system.cpu[2].workload = [Mybench.gobmk] >>> > system.cpu[3].workload = [Mybench.h264ref] >>> > system.cpu[4].workload = [Mybench.bzip2] >>> > system.cpu[5].workload = [Mybench.h264ref] >>> > system.cpu[6].workload = [Mybench.gobmk] >>> > system.cpu[7].workload = [Mybench.libquantum] >>> > >>> > >>> > *************************************************************************************************** >>> > case (b): >>> > >>> > max_insts_all_threads = 100000000 >>> > num of cpus = 8 >>> > num threads per cpu = 1 >>> > system.cpu[0].workload = [Mybench.perlbench] >>> > system.cpu[1].workload = [Mybench.bzip2] >>> > system.cpu[2].workload = [Mybench.h264ref] >>> > system.cpu[3].workload = [Mybench.h264ref] >>> > system.cpu[4].workload = [Mybench.gobmk] >>> > system.cpu[5].workload = [Mybench.libquantum] >>> > system.cpu[6].workload = [Mybench.gobmk] >>> > system.cpu[7].workload = [Mybench.h264ref] >>> > >>> > >>> > Thanks, >>> > >>> > HARI >>> > >>> > _______________________________________________ >>> > 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 >> >> >> >> -- >> Regards >> H A R I >> >> _______________________________________________ >> 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
