mapd2

Central repo for the buildout of MapD V2

Building MapD

MapD uses CMake for its build system. Only the Unix Makefiles and Ninja generators are regularly used. Others, such as Xcode might need some work.

mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE=debug ..
make -j

The following cmake/ccmake options can enable/disable different features:

• -DENABLE_CALCITE=on enable Calcite SQL parser. Default off.
• -DENABLE_RENDERING=on enable backend rendering. Default off.
• -DENABLE_CUDA=off disable CUDA (CUDA still required for build). Default on.
• -DMAPD2_FRONTEND_DOWNLOAD=on download the latest pre-built mapd2-frontend. Default off.
• -DPREFER_STATIC_LIBS=on static link dependencies, if available. Default off.

Dependencies

MapD has the following dependencies:

• CMake 3.3+
• LLVM 3.5
• GCC 4.9+: not required if building with Clang
• Boost 1.5.7+
• Thrift 0.9.2+
• bison++
• Google glog
• Go 1.5+
• OpenJDK
• CUDA 7.0+
• GLEW
• GLFW 3.1.2+
• libpng

//TODO(@vastcharade): add backend rendering deps //TODO(@dwayneberry: add Calcite deps

Generating PDFs of the documentation requires pandoc and texlive (specifically pdflatex).

CentOS 6/7

scripts/mapd-deps-linux.sh is provided that will automatically download, build, and install most dependencies. Before running this script, make sure you have the basic build tools installed:

yum groupinstall "Development Tools"
yum install git

Instructions for installing CUDA are below.

CUDA

scripts/cuda-autoinstall.sh will install CUDA and the drives via the runfile method. For the RPM method, first enable EPEL via yum install epel-release and then follow the instructions provided by Nvidia.

Mac OS X 10.11

scripts/mapd-deps-osx.sh is provided that will automatically install and/or update Homebrew and use that to install all dependencies. Please make sure OSX is completely update to date and Xcode is installed before running.

CUDA

mapd-deps-osx.sh will automatically install CUDA via Homebrew.

Ubuntu 15.10

Most build dependencies are available via APT. Thrift is the one exception and must be built by hand (Thrift 0.9.1 is available in APT, but that version is not supported by MapD).

apt-get update
apt-get install build-essential \
                cmake \
                cmake-curses-gui \
                clang-3.5 \
                clang-format-3.5 \
                llvm-3.5 \
                libboost-all-dev \
                bison++ \
                libgoogle-glog-dev \
                golang \
                libssl-dev \
                libevent-dev \
                libglew-dev \
                libglfw3-dev \
                libpng12-dev

apt-get build-dep thrift-compiler
wget http://apache.claz.org/thrift/0.9.2/thrift-0.9.2.tar.gz
tar xvf thrift-0.9.2.tar.gz
cd thrift-0.9.2
./configure --with-lua=no --with-python=no --with-php=no --with-ruby=no --prefix=/usr/local/mapd-deps
make -j $(nproc)
make install

CUDA

CUDA should be installed via the .deb method, following the instructions provided by Nvidia.

Arch

Assuming you already have yaourt or some other manager that supports the AUR installed:

yaourt -S git cmake boost google-glog extra/jdk8-openjdk clang llvm thrift go cuda nvidia glew glfw libpng
wget https://flexpp-bisonpp.googlecode.com/files/bisonpp-1.21-45.tar.gz
tar xvf bisonpp-1.21-45.tar.gz
cd bison++-1.21
./configure && make && make install

CUDA

CUDA is installed to /opt/cuda instead of the default /usr/local/cuda. You may have to add the following to CMakeLists.txt to support this:

include_directories("/opt/cuda/include")

Environment variables

If using mapd-deps-linux.sh or Ubuntu 15.10, you will need to add following environment variables to your ~/.bashrc or a file such as /etc/profile.d/mapd-deps.sh (or use Environment Modules):

MAPD_PATH=/usr/local/mapd-deps
PATH=$MAPD_PATH/bin:$PATH
LD_LIBRARY_PATH=$MAPD_PATH/lib:$LD_LIBRARY_PATH
LD_LIBRARY_PATH=$MAPD_PATH/lib64:$LD_LIBRARY_PATH
export PATH LD_LIBRARY_PATH

CUDA requires the following environment variables, assuming CUDA was installed to /usr/local/cuda:

CUDA_PATH=/usr/local/cuda
PATH=$CUDA_PATH/bin:$PATH
LD_LIBRARY_PATH=$CUDA_PATH/lib64:$LD_LIBRARY_PATH
export PATH LD_LIBRARY_PATH

CUDA on OS X is usually installed under /Developer/NVIDIA/CUDA-7.5:

CUDA_PATH=/Developer/NVIDIA/CUDA-7.5
PATH=$CUDA_PATH/bin:$PATH
DYLD_LIBRARY_PATH=$CUDA_PATH/lib64:$DYLD_LIBRARY_PATH
export PATH DYLD_LIBRARY_PATH