Skip to content

Latest commit

 

History

History
97 lines (90 loc) · 3.44 KB

COMPILE.md

File metadata and controls

97 lines (90 loc) · 3.44 KB
Raw

Dependencies

Software Directory

First, create a directory where the dependencies will be compiled, e.g.,

mkdir -p ${HOME}/Software

The remainder of this guide assumes such a directory has been created.

METIS

Next, compile the COIN-OR version of METIS, e.g.,

cd ${HOME}/Software
git clone https://github.com/coin-or-tools/ThirdParty-Metis.git
cd ThirdParty-Metis && ./get.Metis
mkdir build
cd build
../configure --prefix=${PWD}
make && make install
export METISDIR=${PWD}

hwloc

Next, ensure the hardware locality library is installed on your system. For example, on Ubuntu 20.04 LTS, this can be accomplished via

sudo apt-get install hwloc libhwloc-dev

BLAS

A number of BLAS libraries can be used for compilation, including the proprietary Intel MKL. One open-source option is OpenBLAS, which can be installed on Ubuntu 20.04 LTS via

sudo apt install libopenblas-dev

This is the library used throughout the remainder of this guide.

CUDA (optional)

If you're installing with NVIDIA CUDA GPU support, ensure it is installed and that the following environment variables are set:

export CUDA_HOME="/usr/local/cuda" # Change this to your system-specific path.
export PATH="${PATH}:${CUDA_HOME}/bin"
export LIBRARY_PATH="${LIBRARY_PATH}:${CUDA_HOME}/lib64"
export LD_LIBRARY_PATH="${LD_LIBRARY_PATH}:${CUDA_HOME}/lib64"
export C_INCLUDE_PATH="${CPLUS_INCLUDE_PATH}:${CUDA_HOME}/include"
export CPLUS_INCLUDE_PATH="${CPLUS_INCLUDE_PATH}:${CUDA_HOME}/include"
export NVCC_INCLUDE_FLAGS="${NVCC_INCLUDE_FLAGS}:-I${CUDA_HOME}/include"

Compilation

Multicore CPUs Only

To compile with only multicore CPU support, execute

cd ${HOME}/Software
git clone https://github.com/ralna/spral.git
cd spral
./autogen.sh # If compiling from scratch.
mkdir build
cd build
CFLAGS=-fPIC CPPFLAGS=-fPIC CXXFLAGS=-fPIC FFLAGS=-fPIC \
   FCFLAGS=-fPIC ../configure --prefix=${PWD}/build \
   --with-blas="-lopenblas" --with-lapack="-llapack" \
   --with-metis="-L${METISDIR}/lib -lcoinmetis" \
   --with-metis-inc-dir="${METISDIR}/include/coin-or/metis"
make && make install

Multicore CPUs and NVIDIA GPUs (optional)

To compile with multicore CPU and NVIDIA GPU support, execute

cd ${HOME}/Software
git clone https://github.com/ralna/spral.git
cd spral
./autogen.sh # If compiling from scratch.
mkdir build
cd build
CFLAGS=-fPIC CPPFLAGS=-fPIC CXXFLAGS=-fPIC FFLAGS=-fPIC \
   FCFLAGS=-fPIC NVCCFLAGS="-shared -Xcompiler -fPIC" \
   ../configure --prefix=${PWD}/build \
   --with-blas="-lopenblas" --with-lapack="-llapack" \
   --with-metis="-L${METISDIR}/lib -lcoinmetis" \
   --with-metis-inc-dir="${METISDIR}/include/coin-or/metis"
make && make install

If your GPUs are not being recognized, consider uncommenting line 91 and commenting line 92 of src/hw_topology/hwloc_wrapper.hxx, then recompiling. The methods for recognizing GPUs do not seem to function independently of the type of system being used.

Usage

For future use, set the SPRAL directory environment variable via

export SPRALDIR=${PWD}/build

from the ${HOME}/Software/spral directory. Also, ensure the following environment variables are set when using the library:

export OMP_CANCELLATION=TRUE
export OMP_PROC_BIND=TRUE