Hi Abdlerhman, One possible explanation is as follows: You are running the same program on all cores with the m5 checkpoint call and then you take at maximum one checkpoint, which will be taken by whichever core hits the checkpoint call first. In the first two cases you mentioned (Atomic vs. O3 CPU), the core which takes the checkpoint could be different in both cases (and in addition all the cores could be in a different state), and that means the system state at the point of restore could be different which can lead to difference in statistics you are observing. Please, note that the Atomic CPU does not model timing of cache/memory accesses (in contrast to O3 CPU) and as a result you might not be able to see the effects of shared resources (e.g. LLC).
-Ayaz On Fri, Mar 24, 2023 at 4:07 PM Abdlerhman Abotaleb via gem5-users < gem5-users@gem5.org> wrote: > I have a program bin.riscv that is running on 4 prcoessors simulatenously. > Inside riscv.bin: > // some code > m5_checkpoint(0,0); > m5_reset_stats(0,0); > // some code > I'm running GEM5 using the following commands: > > 1. First > gem5.opt ./configs/example/se.py -n 4 --caches --max-checkpoints 1 > --cpu-type AtomicSimpleCPU --cmd 'a.riscv;a.riscv;a.riscv;a.riscv' > > 2. Then > gem5.opt ./configs/example/se.py -n 4 --caches --checkpoint-restore=1 > --restore-with-cpu DerivO3CPU --cmd 'a.riscv;a.riscv;a.riscv;a.riscv' > > The statistics are different than if I run the same code (including > writing checkpints and reset stats) but with the following gem5 command: > > 1. gem5.opt ./configs/example/se.py -n 4 --caches --cpu-type > DerivO3CPU --cmd 'a.riscv;a.riscv;a.riscv;a.riscv' > > > It is worth mentioning that also the different CPUs don't have the same > number of instructions instsIssued or numInsts (i.e committed instructions) > (from GEM5 generated stats file) . > > Do you have explanations or any useful suggestions? > Thanks. > > > > _______________________________________________ > gem5-users mailing list -- gem5-users@gem5.org > To unsubscribe send an email to gem5-users-le...@gem5.org >
_______________________________________________ gem5-users mailing list -- gem5-users@gem5.org To unsubscribe send an email to gem5-users-le...@gem5.org
