Hi,

I've run multi-threaded applications in fs/se simulation. But I found that
the simulator cannot go over the omp_barrier/pthread_barrier. The tracing
results show that the cpus repetitively access the same set of addresses
(hitting in cache), and dram controllers receive no requests. This is very
like the behavior of spin lock.

I notice such behaviors for both timing and detailed cpu models during fs
and se simulations. The atomic cpu model behaves normally. I wonder whether
the problem is in the execution of atomic instructions that needed by
barrier.

Has anyone noticed the same problem?

thanks,
Zehan
_______________________________________________
gem5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users

Reply via email to