Hi All,
I am trying to run a full system simulation of X86 architecture. I've tried
different configurations
of fs and ruby_fs with timing and detailed CPU models. These are examples
that are examined:

build/X86_MESI_CMP_directory/gem5.opt configs/example/ruby_fs.py --disk=...
/x86root-parsec.img --num-cpus=16 --num-dirs=16 --topology=Mesh
--mesh-rows=4 --garnet-network=fixed --num-l2caches=16 --l1d_size="32kB"
--l1i_size="32kB" --l2_size="1024kB"  --cpu-type=timing

or

build/X86_MOESI_hammer/gem5.opt  configs/example/ruby_fs.py
--disk=.../x86root-parsec.img --num-cpus=16 --num-dirs=16 --topology=Mesh
--mesh-rows=4 --num-l2caches=16 --l1d_size="32kB" --l1i_size="32kB"
--l2_size="1024kB"  --cpu-type=detailed

or

build/X86_MOESI_hammer/gem5.opt configs/example/fs.py
--disk=.../x86root-parsec.img --num-cpus=16 --num-dirs=16 --num-l2caches=16
--l1d_size="32kB" --l1i_size="32kB" --l2_size="1024kB"  --cpu-type=detailed

Then, I tried to get a checkpoint (preferably with timing mode CPU) and
then restore them and switch CPU model to detailed mode, or checkpoint it
directly in detailed mode. For example:

build/X86_MOESI_hammer/gem5.opt configs/example/ruby_fs.py
--disk=.../x86root-parsec.img --num-cpus=16 --num-dirs=16 --topology=Mesh
--mesh-rows=4 --num-l2caches=16 --l1d_size="32kB" --l1i_size="32kB"
--l2_size="1024kB"  --restore-with-cpu=detailed -r 1

The problem is that either if  I take a checkpoint at timing mode CPU and
try to switch detailed mode at restore time, or I checkpoint it directly in
detailed mode cause a huge memory leak, and simulation would quickly
consume all memory and swap spaces that leads to killing it by OS. I could
not find any reason for it!

Did anyone have a similar issue? I would very appreciate your quick answer
to it.

Thanks
-Yasha
_______________________________________________
gem5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users

Reply via email to