Hello
I know this was discussed before but I did not find anybody who actually make
it work. Just like others before me I am facing difficulties in implementing
the back and forth switch.
What I did so far:
- I added the following to the configs/common/Simulation.py (the red code is
what I added)
...............................................................................................................................................................
## Adjust max tick based on our starting tick
maxtick = maxtick - int(cpts[cpt_num - 1])
checkpoint_dir = joinpath(cptdir, "cpt.%s" % cpts[cpt_num - 1])
simple_cpu_class = AtomicSimpleCPU
detailed_cpu_class = DerivO3CPU
detailed_switch_cpus = [detailed_cpu_class(defer_registration=True,
cpu_id=(np+i)) for i in xrange(np)]
simple_switch_cpus = [simple_cpu_class(defer_registration=True,
cpu_id=(2*np+i)) for i in xrange(np)]
for i in xrange(np):
detailed_switch_cpus[i].system = testsys
simple_switch_cpus[i].system = testsys
if not buildEnv['FULL_SYSTEM']:
detailed_switch_cpus[i].workload = testsys.cpu[i].workload
simple_switch_cpus[i].workload = testsys.cpu[i].workload
detailed_switch_cpus[i].clock = testsys.cpu[0].clock
simple_switch_cpus[i].clock = testsys.cpu[0].clock
testsys.switch_detailed = detailed_switch_cpus
testsys.switch_simple = simple_switch_cpus
switch_cpu_list = [(testsys.cpu[i], detailed_switch_cpus[i] ) for i in
xrange(np)]
switch_cpu_list1 = [(detailed_switch_cpus[i], simple_switch_cpus[i] ) for i
in xrange(np)]
switch_cpu_list2 = [(simple_switch_cpus[i], detailed_switch_cpus[i] ) for i
in xrange(np)]
m5.instantiate(checkpoint_dir)
...............................................................................................................................................................
else: # no checkpoints being taken via this script
if options.fast_forward:
m5.stats.reset()
print "**** REAL SIMULATION ****"
exit_event = m5.simulate(maxtick)
exit_cause = exit_event.getCause()
if exit_cause == "switchcpu":
print "Switchcpus Atomic -> Detailed"
m5.stats.dump()
m5.stats.reset()
m5.changeToTiming(testsys)
m5.drain(testsys)
m5.switchCpus(switch_cpu_list)
m5.resume(testsys)
exit_event = m5.simulate(maxtick-m5.curTick())
exit_cause = exit_event.getCause()
while exit_cause == "switchcpu":
print "Switchcpus Detailed -> Atomic"
m5.stats.dump()
m5.stats.reset()
m5.changeToAtomic(testsys)
m5.drain(testsys)
m5.switchCpus(switch_cpu_list1)
m5.resume(testsys)
exit_event = m5.simulate(maxtick-m5.curTick())
exit_cause = exit_event.getCause()
if exit_cause == "switchcpu":
print "Switchcpus Atomic -> Detailed"
m5.stats.dump()#"Atomic")
m5.stats.reset()
m5.changeToTiming(testsys)
m5.drain(testsys)
m5.switchCpus(switch_cpu_list2)
m5.resume(testsys)
exit_event = m5.simulate(maxtick-m5.curTick())
exit_cause = exit_event.getCause()
while exit_event.getCause() == "checkpoint":
m5.checkpoint(joinpath(cptdir, "cpt.%d"))
...............................................................................................................................................................
- I modified the BaseCPU::takeOverFrom function in src/cpu/base.cc so that port
assignment is performed even if the newCPU was previously connected in the
system:
...............................................................................................................................................................
if (1){
Port *peer = oldCPU->getPort("icache_port")->getPeer();
ic->setPeer(peer);
peer->setPeer(ic);
}
if (1) {
Port *peer = oldCPU->getPort("dcache_port")->getPeer();
dc->setPeer(peer);
peer->setPeer(dc);
}
...............................................................................................................................................................
I start the simulation with AtomicSimpleCPU model:
./build/ALPHA_FS/gem5.opt --outdir=test configs/example/fs.py -n 2 --caches
--l2cache
I trigger the switches from terminal and the first 2 switches (first pass in
the while loop in Simulation.py) works fine, but after that the simulation
appears to hang. I ran it through GDB and it appears that the simulation is
still running but is running extremely slow. I can't figure it out what is
going wrong. Am I approaching this the wrong way? Am I doing something wrong
when I am setting up the switching environment? Any suggestions?
Thanks in advance!
Alexandru
- I modified the
_______________________________________________
gem5-users mailing list
gem5-users@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
