Skip to content

Latest commit

 

History

History
101 lines (60 loc) · 3.79 KB

2019-01-09-Install and Run GPGPUSim.md

File metadata and controls

101 lines (60 loc) · 3.79 KB
Raw
layout comments title categories
post
true
Install and Run GPGPU-Sim
GPU Machine-Learning Simulator

GPGPU-Sim is a cycle-level simulator for modeling contemporary GPUs running CUDA and OpenCL workloads. The current GPGPU-Sim supports the GPU simulation with four kinds of architectures, i.e., GTX480, QuadroFX5600, QuadroFX5800, and TeslaC2050 architectures. This blog introduces the detailed steps to install and run GPGPU-Sim.

1 Download and Install NVDIA CUDA 4.0

GPGPU-Sim has to be run with NVDIA CUDA and does not support the CUDA versions larger than 4.0. Hence, we should first install NVDIA CUDA 4.0. The linux OS in my computer is Ubuntu 18.04, and the gcc version is 7.3.0. To install NVDIA CUDA 4.0, please do the following setps.

1) Download the CUDA Toolkit for Ubuntu Linux 10.10 and GPU Computing SDK code samples from the NVDIA website.

2) Install CUDA Toolkit for Ubuntu Linux 10.10 first:

chmod +x cudatoolkit_4.0.17_linux_64_ubuntu10.10.run
sudo ./cudatoolkit_4.0.17_linux_64_ubuntu10.10.run

The CUDA Toolkit has been installed in the path of /usr/local/cuda in default.

  1. Add the path of CUDA Toolkit into the ~/.bashrc file:

    echo 'export PATH=$PATH:/usr/local/cuda/bin' >> ~/.bashrc echo 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib:/usr/local/cuda/lib64' >> ~/.bashrc source ~/.bashrc

  2. Install GPU Computing SDK code samples:

    chmod +x gpucomputingsdk_4.0.17_linux.run sudo ./gpucomputingsdk_4.0.17_linux.run

The GPU Computing SDK has been installed in the path of ~/NVIDIA_GPU_Computing_SDK in default.

  1. Install gcc-4.4 and g++-4.4 (since CUDA 4.0 supports the gcc version until 4.4):

    apt-get install gcc-4.4 g++-4.4

If the error package gcc-4.4 is not available, but is referred to by another package occurs, do the following steps to address it:

vim /etc/apt/sources.list

Add the two-line codes into the opened file:

deb http://dk.archive.ubuntu.com/ubuntu/ trusty main universe    
deb http://dk.archive.ubuntu.com/ubuntu/ trusty-updates main universe

Then, update the apt source:

apt-get update

By now, the gcc-4.4 and g++-4.4 have been installed.

  1. Change the gcc/g++ in the system to gcc-4.4/g++4.4:

    sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-7 150 sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-4.4 100 sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-7 150 sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-4.4 100

Select the 4.4 version by using update-alternatives:

sudo update-alternatives --config gcc
sudo update-alternatives --config g++

2 Download and Install GPGPU-Sim

  1. Download GPGPU-Sim from GitHub

    git clone https://github.com/gpgpu-sim/gpgpu-sim_distribution.git

  2. Install dependencies

    sudo apt-get install build-essential xutils-dev bison zlib1g-dev flex libglu1-mesa-dev sudo apt-get install doxygen graphviz sudo apt-get install python-pmw python-ply python-numpy libpng12-dev python-matplotlib sudo apt-get install libxi-dev libxmu-dev freeglut3-dev

  3. Add the CUDA_INSTALL_PATH into the ~/.bashrc file:

    echo 'export CUDA_INSTALL_PATH=/usr/local/cuda' >> ~/.bashrc source ~/.bashrc

  4. Build GPGPU-Sim:

    make

During the building, if there is an error cuobjdump.l:110: error: unterminated comment cuobjdump.l:108: error: expected declaration or statement at end of input, remove the comments in cuobjdump.l:108-109.

  1. Run GPGPU-Sim:

Copy the contents of a GPU config, e.g., configs/GTX480/*, to your application's working directory, and then run a CUDA application.

mkdir test
cd test/
cp ../configs/GTX480/* ./