Just to clarify, "deadlock" in this case doesn't necessarily mean that the application you're simulating deadlocks. More likely (especially in this case) it's a deadlock in the simulator itself, e.g., some corner case where the CPU is waiting for something and due to a bug the event that's supposed to wake it up doesn't get scheduled or gets ignored or something like that.
Steve On Tue, Jun 15, 2010 at 4:26 AM, Korey Sewell <ksew...@umich.edu> wrote: > What is the expected number of instructions for your simulation? > What is the number in the stats file after it exits? > > The answers to those questions will help you find where the deadlock point > potentially is. > > Also, if you turn the progress intervals on you can see where the deadlock > potentially is at well because you start seeing intervals of 0 instructions. > (--prog_intvl=X on the command line.) > > On Tue, Jun 15, 2010 at 3:13 AM, Maximilien Breughe > <maximilien.breu...@elis.ugent.be> wrote: >> >> Hi all, >> >> After executing a subset of MiBench I received this exit cause for a few >> benchmarks: >> Exiting @ cycle 9223372036854775807 because simulate() limit reached >> >> CRC32 and gsm encode with the large input set are the two benchmarks that >> cause this message. >> >> On the mailing list I read that this happens due to deadlocking for >> example. >> However, I run this code on a single InOrderCPU and I don't see where CRC >> could possibly deadlock. Besides, it executes correctly on the Atomic CPU. >> >> Does anybody have ideas about possible causes? Or did anybody ran >> CRC32/gsm encode large on the InOrderCPU to completeness? >> >> >> Regards, >> >> Max >> _______________________________________________ >> m5-users mailing list >> m5-users@m5sim.org >> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users > > > > -- > - Korey > > _______________________________________________ > m5-users mailing list > m5-users@m5sim.org > http://m5sim.org/cgi-bin/mailman/listinfo/m5-users > _______________________________________________ m5-users mailing list m5-users@m5sim.org http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
