mc2lib_for_gem5

Cloning the REPO

We use gitHub's git-lfs support to store several large binary files. After cloning the repo, please do a git lfs fetch and update from the repository root folder to retrieve them.

git lfs fetch
git lfs pull

Install GEM5 instructions:

• To build gem5, we need to install the following softwares:sudo apt-get install mercurial scons swig gcc m4 python python-dev libgoogle-perftools-dev g++ zlib-dev zlib1g-dev
• Note that for g++ we used the version 4.8.
• Get gem5 from gem5 website
• Get the patch from this link
• Patch gem5 with Mcversi patch : cd gem5 , patch < mcversi_gem5.patch
• Note that we copy the files generated in the gem5 directory (mcversi.* files) after patching the gem5 to the gem5/src/sim directory in order to fix some errors we got, In addition we load the mc2lib directory from mc2lib to the gem5/src/sim directory.
• Modify some files to set Mcversi paper parameters (the system parameters in the paper ):

1. ROB entries = 40: modify the default value of ROBEntries in src/cpu/o3/O3CPU.py. The default value is 192; we modify it to 40.
2. LSQ entries = 32: modify the default value of LQEntries and SQEntries in src/cpu/o3/O3CPU.py. Since gem5 has a separate queue for loads and stores, we set each variable to 16 so that the aggregated size becomes 32.
3. L1 and L2 hit latency: modify the default value of hit_latency in the L1Cache class (to 3) and L2Cache class (to 30) in configs/common/Caches.py.
4. We still need to know how to set the Memory latency, so we supposed that we use the default value.

Build the X86 Architecture:

• Build the arch x86 : scons build/X86/gem5.opt -j8 with –j is the number of cores to create.
• To use full system:
  • Create full_system_images directory: mkdir full_system_images, cd full_system_images,wget http://www.m5sim.org/dist/current/x86/x86-system.tar.bz2and wget http://www.m5sim.org/dist/current/x86/config-x86.tar.bz2 (source : http://gem5.org/Download)
  • Modify : gem5/configs/common/SysPaths.py to set: path = [ ’/dist/m5/system’, ’’ ] . Our case: path = [ '/dist/m5/system', '/home/rachid/gem5/full_system_images/']
  • AND gem5/configs/common/Benchmarks.py. There we specify the name of our linux image: return env.get(’LINUX_IMAGE’, disk(’linux-parsec-2-1-m5.img’))
• Add alpha disk image and binaries to full_system_images/: Alpha (to avoid some errors we get: can't find linux-bigswap2.img error)

McVerSi Guest Workload:

• In order to run mcversi under x86 we need to copy the guest workload ( we used a pre-compiled guest workload, file: guest_workload.x86-64) to x86 image following this steps:
  1. Extract the file from mcversi_guest_workload_gem5.tar.gz
  2. Make a directory called mount inside full_system_images directory: mkdir mount
  3. Mount the image, type from full_system_images directory: sudo mount -o loop,offset=32256 disks/linux-x86.img ./mount
  4. cd mount, sudo mkdir movefile
  5. copy the file to image: sudo cp /path_to_file_guest_workload.x86-64 /movefile
  6. Type whether successful by cd movefile and ls
  7. Unmount the image by typing: sudo umount mount from full_system_images directory.

Taking a checkpoint and restoring it:

• Start up the architecture with cpu-type=timing : ./build/X86/gem5.opt configs/example/fs.py --disk-image=/home/rachid/gem5/full_system_images/disks/linux-x86.img --kernel=/home/rachid/gem5/full_system_images/binaries/x86_64-vmlinux-2.6.22.9.smp --mem-size=512MB --cpu-type=timing --num-cpus=8 --cpu-clock=2GHz --ruby --caches --l2cache --l1d_size=32kB --l1i_size=32kB --cacheline_size=64 --l1i_assoc=4 --l1d_assoc=4 --num-l2caches=8 --l2_size=128kB --l2_assoc=4 --garnet-network=fixed --topology=Mesh --mesh-rows=2

• Take a checkpoint m5 checkpoint

• Restore the checkpoint: ./build/X86/gem5.opt configs/example/fs.py -r N OR ./build/X86/gem5.opt configs/example/fs.py --checkpoint-restore=N , The number N is integer that represents checkpoint number which usually starts from 1 then increases incrementally to 2,3,4...

• Restore the checkpoint with specific CPU type (for instance wpu type: detailed): ./build/X86/gem5.opt configs/example/fs.py -r N --restore-with-cpu=detailed

Note: By default, gem5 assumes that the checkpoint is to be restored using Atomic CPUs. This may not work if the checkpoint was recorded using Timing / Detailed / Inorder CPU. One can mention the option --restore-with-cpu on the command line. The cpu type supplied with this option is then used for restoring from the checkpoint.

Run McVerSi Guest Workload under X86 architecture:

• Start up the full system with the paper parameters: ./build/X86/gem5.opt configs/example/fs.py --disk-image=/home/rachid/gem5/full_system_images/disks/linux-x86.img --kernel=/home/rachid/gem5/full_system_images/binaries/x86_64-vmlinux-2.6.22.9.smp --mem-size=512MB --cpu-type=detailed --cpu-clock=2GHz --ruby --caches --l2cache --num-cpus=4 --l1d_size=32kB --l1i_size=32kB --cacheline_size=64 --l1i_assoc=4 --l1d_assoc=4 --num-l2caches=8 --l2_size=128kB --l2_assoc=4 --topology=Mesh --mesh-rows=2 --num-dirs=18 --garnet-network=fixed
• Use telnet to access to the console: telnet localhost 3456
• cd movefile :
• Run the pre-compiled McVerSi Guest Workload using provided parameters in this link with 4 threads instead 10 for the first time: ./guest_workload 4 0xf00200 0x09140010 0x100000000
• Run McVerSi Guest Workload with 10 threads: ./guest_workload 10 0xf00200 0x09140010 0x100000000