Hi there,
I want to do a warm up before starting the real data-collection
simulation. I tried to do that by calling the
g_system_ptr->clearStats(), and the benchmark that I'm now using is SPEC
OpenMP.
But the strange thing is that I can run the simulation using the
original GEMS and simics, after my modification, some of the bechmark
runs into invalid transition of the directory while some benchmark end
up with segmentation fault.
Does anyone know about the clearStats function? The clearStats of
Profiler seems to be a little bit complex, but that is not supposed to
incur this weired program behavior, right? Does anyone have any idea
about what is going wrong?
Thanks in advance,
Guoqiang Yang
|