Hi Giacomo, Thanks for pointing it out, it helps a lot in defining the problem, although I think it needs much more effort than I was thinking. So, the problem here is that, when you already restoring from a checkpoint, and then drop checkpoints, the cpu are now called testsys.switch_cpu, you will see the _pid under testsys.switch_cpu section, but not under testsys.cpu section. But later, where you restore again from these checkpoint, unserialize method still are looking for testsys.switch_cpu section.
Though, you can copy paste the _pid (specifically in this case, the reported error), go pass that, there are more fields/sections missing. For insurance, the testsys.cpu does not have the testsys.cpu.xc.0 whole section, (not sure if you can copy paste whole section). Even you can copy paste everything that’s missing and assume they are correct, it may not be the elegant solution. Change how the checkpoints dropped/restore seems more reasonable to me, but for sure takes more effort. Correct me if I am wrong about above or missing something. For now, I think I will just pass this issue and not use these checkpoints. Best, Heng > On Mar 19, 2020, at 04:18, Giacomo Travaglini <[email protected]> > wrote: > > It probably happens that when you try to restore (unserialize) the > checkpoint, The BaseCPU::unserialize method in src/cpu/base.c is trying to > restore the _pid field, but id doesn't find it in the checkpoint. > > If you take a look at the checkpoint you will see that under cpu0 it is > probably missing this field. > Which means that a previous serialization didn't write the _pid field for > whatever reason. > > If you debug it with gdb you can see why this didn't happen > > Giacomo > > -----Original Message----- > From: gem5-users <[email protected]> On Behalf Of HENG ZHUO > Sent: 19 March 2020 06:15 > To: gem5 users mailing list <[email protected]> > Subject: [gem5-users] Cannot restoring from fs HPI checkpoint > > Hi, > > I boot up the kernel and do preparation using AtomicSimpleCPU, then restore > with HPI to run the workload, this part was fine. Then, when tuning the > actual workload, I was using —take-checkpoint to drop checkpoints. However, I > was not able to restore from these checkpoints. Commands used as: > > 1)Boot up the kernel used: > --outdir=/gem5/m5out/masstree/100/ configs/example/fs.py > --machine-type=VExpress_EMM64 --cpu-type=AtomicSimpleCPU --num-cpus\ > =4 --cpu-clock=2GHz --sys-clock=2GHz > --kernel=/images/arm2018/binaries/vmlinux.vexpress_emm64 > --dtb-filename=/images/arm2018/bin\ > aries/armv8_gem5_v1_4cpu.dtb -b masstree > > 2)Restore from boot up, running the workload as: > --outdir=/gem5/m5out/masstree/100/ configs/example/fs.py > --machine-type=VExpress_EMM64 --cpu-type=HPI --num-cpus=4 --cpu-clock=2\ GHz > --sys-clock=2GHz --kernel=/images/arm2018/binaries/vmlinux.vexpress_emm64 > --dtb-filename=/images/arm2018/bin\ > aries/armv8_gem5_v1_4cpu.dtb -b masstree --caches --l2cache --l1i_size=32kB > --take-checkpoint=13000000000000,1000000000000 -r 3 > > 3)Restore form checkpoints makes during running with workload(FS, HPI): > --outdir=/m5out/masstree/100/ configs/example/fs.py > --machine-type=VExpress_EMM64 --cpu-type=HPI --num-cpus=4 --cpu-clock=2\ GHz > --sys-clock=2GHz --kernel=/images/arm2018/binaries/vmlinux.vexpress_emm64 > --dtb-filename=/images/arm2018/bin\ > aries/armv8_gem5_v1_4cpu.dtb -b masstree --caches --l2cache --l1i_size=32kB > -r 8 —restore-with-cpu=HPI > > Step 3 is where I failed with this error: > ……. > warn: No functional unit for OpClass SimdPredAlu > warn: Sockets disabled, not accepting gdb connections > warn: optional parameter testsys.cpu0:lastStopped not present > fatal: Can't unserialize 'testsys.cpu0:_pid' > Memory Usage: 4929176 KBytes > > > Already tried with no —resotre-with-cpu option, no luck with that. Any help > is appreciated. > > Thanks, > Heng > _______________________________________________ > gem5-users mailing list > [email protected] > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users > IMPORTANT NOTICE: The contents of this email and any attachments are > confidential and may also be privileged. If you are not the intended > recipient, please notify the sender immediately and do not disclose the > contents to any other person, use it for any purpose, or store or copy the > information in any medium. Thank you. > _______________________________________________ > gem5-users mailing list > [email protected] > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users _______________________________________________ gem5-users mailing list [email protected] http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
