This document includes instructions on how to create an Ubuntu 18.04 disk-image with PARSEC benchmark installed. The disk-image will be compatible with the gem5 simulator.
This is how the
src/parsec-tests/ directory will look like if all the artifacts are created correctly.
parsec/ |___ gem5/ # gem5 folder | |___ disk-image/ | |___ build.sh # the script downloading packer binary and building the disk image | |___ shared/ | |___ parsec/ | |___ parsec-image/ | | |___ parsec # the disk image will be here | |___ parsec.json # the Packer script | |___ parsec-install.sh # the script to install PARSEC | |___ post-installation.sh # the script to install m5 | |___ runscript.sh # script to run each workload | |___ parsec-benchmark # the parsec benchmark suite | |___ configs | |___ system # system config directory | |___ run_parsec.py # gem5 run script | |___ configs-mesi-two-level | |___ system # system config directory | |___ run_parsec_mesi_two_level.py # gem5 run script | |___ README.md
Notice that there are two sets of system configuration directories and run scripts. For further detail on the config files look here.
In order to build the disk-image for PARSEC tests with gem5, build the m5 utility in
src/parsec-tests/ using the following:
git clone https://gem5.googlesource.com/public/gem5 cd gem5/util/m5 scons build/x86/out/m5
We use packer to create our disk-image. The instructions on how to install packer is shown below:
cd disk-image ./build.sh # the script downloading packer binary and building the disk image
In order to build the disk-image first the script needs to be validated. Run the following command to validate
./packer validate parsec/parsec.json
After the script has been successfuly validated you can create the disk-image by runnning:
./packer build parsec/parsec.json
You can find the disk-image in
There are two sets of run scripts and system configuration files in the directory. The scripts found in
configs use the classic memory system while the scripts in
configs-mesi-two-level use the ruby memory system with MESI_Two_Level cache coherency protocol. The parameters used in the both sets of experiments are explained below:
x264. For more information on the workloads can be found at https://parsec.cs.princeton.edu/.
configs, the only valid option is
1. When using
configs-mesi-two-levelthe number of supported cpus is show in the table below:
|CPU Model||Core Counts|
Below are the examples of running an experiment with the two configurations.
<gem5 X86 binary> configs/run_parsec.py <kernel> <disk> <cpu> <benchmark> <size> <num_cpus> <gem5 X86_MESI_Two_Level binary> configs-mesi-two-level/run_parsec.py <kernel> <disk> <cpu> <benchmark> <size> <num_cpus>
The working status of PARSEC runs for gem5-20 has been documented here.