On Fri, 1 Jun 2012, Volos Stavros wrote:
Hi Nilay,
I didn't receive the email in person. I just read your reply in the digest
email. Thanks for your reply.
A brief overview of the error is as follows:
Step A
---------
I used the following command to restore a checkpoint that was obtained using
the fast mode. I executed only 1 M instructions and
then took a ruby checkpoint.
./build/ALPHA_MOESI_hammer/gem5.opt -d
/home/volos/volos_icfiler3/gem5_checkpoints/parsec/bodytrack_new
configs/example/ruby_fs.py -n 2 -I 1000000 --checkpoint-restore=1
--l1i_size=32kB --l1d_size=32kB --l2_size=4MB --num-l2caches=2
--topology=Crossbar --restore-with-cpu=timing --checkpoint-at-end
Output:
info: Entering event queue @ 2573654899500. Starting simulation...
Writing checkpoint
info: Entering event queue @ 2573654901000. Starting simulation...
Looking at the folder of the checkpoint, indeed the checkpoint is written
correctly:
ls ~/volos_icfiler3/gem5_checkpoints/parsec/bodytrack_new/cpt.2573654901000/
m5.cpt system.disk0.image.cow system.disk2.image.cow system.physmem.physmem
system.ruby.cache.gz system.ruby.memory.gz
What worries me about the output is that although the simulation should have
exited after writing the checkpoint, it continued for
Step B
---------
Next, I tried to restore the above checkpoint using the following command:
./build/ALPHA_MOESI_hammer/gem5.opt -d
/home/volos/volos_icfiler3/gem5_checkpoints/parsec/bodytrack_new
configs/example/ruby_fs.py -n 2 -I 1000000 --checkpoint-restore=2
--l1i_size=32kB --l1d_size=32kB --l2_size=4MB --num-l2caches=2
--topology=Crossbar --cpu-type=detailed
However, I got the following error:
gem5.opt: build/ALPHA_MOESI_hammer/mem/ruby/system/MemoryVector.hh:227: void
MemoryVector::populatePages(uint8*): Assertion `m_pages[i] == 0' failed.Program
aborted at cycle 2573654901000
Aborted
It seems that it cannot load the checkpoints properly. Am I doing something
wrong?
Regards and thanks,
-Stavros.
On Jun 1, 2012, at 1:42 AM, wrote:
Dear all,
I obtained the latest version of gem5-dev and I was able to run simulations
with various gem5 modes.
I want to create a ruby checkpoint using the TimingSimpleCPU and then load it
to perform OoO simulations.
I am able to create a ruby checkpoint. However, I cannot load it to perform a
Ruby+OoO simulation.
Does gem5 support loading a ruby checkpoint to perform a Ruby+OoO simulation?
Thanks in advance,
-Stavros.
I am aware of this issue. You can look at this thread for an explanation -
http://www.mail-archive.com/gem5-users@gem5.org/msg03379.html
There are a couple of things that you can do here --
a. Comment out the call to setupFuncEvents() in function
AlphaSystem::loadState.
b. Try using the gem5.fast binary instead of gem5.opt.
--
Nilay
_______________________________________________
gem5-users mailing list
gem5-users@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
