When we do the checkingpoint and resume later on same type of CPU, does the
total execution cycles time overhead come from the cache and other resource
warm up? From my understanding, the checkingpoint will not save those
status information.
I did an experiment on this. I run a benchmark program in two steps. After
some clock cycles create a checking point, then load that checking point
from the previous run and continue.
Then I take a look of the "system.cpu.numCycles" in stats.txt file from
these two steps.
If I sum up them up, it is interesting sometimes the total will be smaller
than the single run number of cycles. How could this happen? Thanks
build/ALPHA_MOESI_hammer/gem5.fast configs/example/se.py \
--cpu-type=detailed --caches --l2cache --l1i_size=32kB \
--l1d_size=32kB --l2_size=512kB -c ./benchmark/prog \
-I 1400000 --checkpoint-dir m5out --checkpoint-at-end
build/ALPHA_MOESI_hammer/gem5.fast configs/example/se.py \
--cpu-type=detailed --caches --l2cache --l1i_size=32kB \
--l1d_size=32kB --l2_size=512kB -c ./benchmark/prog \
--checkpoint-dir m5out -r 1 --restore-with-cpu detailed
_______________________________________________
gem5-users mailing list
gem5-users@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
