I am running ocean with small inputs. If I just have one "process" which is an ocean benchmark and run it on all cpus it goes through. Now when i create 2 processes "process1" and "process2" and both of them are still the same ocean benchmark I get this error.
So in effect I have separated the definitions of processes but they are still doing the same thing. ----- Original Message ----- From: "Steve Reinhardt" <[email protected]> To: "M5 users mailing list" <[email protected]> Sent: Tuesday, May 26, 2009 11:17:21 PM GMT -08:00 US/Canada Pacific Subject: Re: [m5-users] m5 with Multiprogramming workloads What applications are you running? If they're calling sys_clone then they're trying to create more threads, which is why I assumed they are multithreaded programs. On Tue, May 26, 2009 at 11:15 PM, Siddharth Gupta < [email protected] > wrote: I give num_cpus = 4 and create 2 processes. 2 cpus run one process. How is this creating less number of threads than is needed. Each cpu would just need one thread to run the process it has been assigned. Am I missing something here? Siddharth ----- Original Message ----- From: "Steve Reinhardt" < [email protected] > To: "M5 users mailing list" < [email protected] > Cc: "prasanthi" < [email protected] > Sent: Tuesday, May 26, 2009 10:58:54 PM GMT -08:00 US/Canada Pacific Subject: Re: [m5-users] m5 with Multiprogramming workloads If you're using a multithreaded workload, you have to make sure you have enough CPUs for all the threads that get created. So if you're giving the process half as many CPUs you need to tell it to create half as many threads. On Tue, May 26, 2009 at 10:32 PM, Siddharth Gupta < [email protected] > wrote: Hi, I had one more question- I also tried to run different processes on different cpus but it gives the following error - fatal: Called sys_clone, but no unallocated thread contexts found. I created process1 and process2 instead of the usual process and then defined each odd cpu workload as process1 and each even cpu workload as process2. If I just define all of them as process1 the error goes away. Any suggestions? Thanks, Siddharth Gupta ----- Original Message ----- From: "Siddharth Gupta" < [email protected] > To: "M5 users mailing list" < [email protected] > Cc: "prasanthi" < [email protected] > Sent: Tuesday, May 26, 2009 8:38:41 PM GMT -08:00 US/Canada Pacific Subject: Re: [m5-users] m5 with Multiprogramming workloads Hi, Thanks for the update. I tried to run multiple threads concurrently using SMT mode with the structure present in the se.py file. I am now getting the following error. build/SPARC_SE/cpu/o3/cpu.cc:278: FullO3CPU<Impl>::FullO3CPU(DerivO3CPUParams*) [with Impl = O3CPUImpl]: Assertion `params->numPhysIntRegs >= numThreads * SparcISA::NumIntRegs' failed. Any help would be great. Thanks, Siddharth ----- Original Message ----- From: "Lisa Hsu" < [email protected] > To: "M5 users mailing list" < [email protected] > Cc: "prasanthi" < [email protected] > Sent: Tuesday, May 26, 2009 5:44:07 PM GMT -08:00 US/Canada Pacific Subject: Re: [m5-users] m5 with Multiprogramming workloads Heh. Me neither, obviously. Sorry about that. However, the mailing list archive is definitely helpful, there are a lot of posts on this, some with sample bits of config script code. Lisa On Tue, May 26, 2009 at 5:38 PM, Steve Reinhardt < [email protected] > wrote: Sorry, I didn't realize the relevant FAQ entry was so singularly unhelpful. Not sure what happened there. I've just updated it: http://m5sim.org/wiki/index.php/Frequently_Asked_Questions#How_do_I_run_multiprogram_workloads_on_M5.3F On Tue, May 26, 2009 at 5:28 PM, Siddharth Gupta < [email protected] > wrote: Hi Steve, Thanks for your reply. Could you give us some pointers on how this can be done. Or point us to the right document. We haven't been able to find this. Siddharth ----- Original Message ----- From: "Steve Reinhardt" < [email protected] > To: "M5 users mailing list" < [email protected] > Cc: "prasanthi" < [email protected] > Sent: Tuesday, May 26, 2009 5:23:29 PM GMT -08:00 US/Canada Pacific Subject: Re: [m5-users] m5 with Multiprogramming workloads Yes. On Mon, May 25, 2009 at 4:52 PM, Siddharth Gupta < [email protected] > wrote: Hi, We are trying to run M5 with multiprogramming workloads so that we can stress the last level cache. Is this possible in M5. Can we schedule one benchmark on one CPU and another on another CPU? Its not necessary for us to be able to do multithreading. Please let us know. Siddharth Gupta _______________________________________________ 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 _______________________________________________ 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 _______________________________________________ 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 _______________________________________________ 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
