Gitingest

Include files under: 50kB

Private Repository NEW

Summary

Repository: intel/torch
Files analyzed: 33

Estimated tokens: 53.1k

Download

Directory Structure

Directory structure:

└── intel-torch/

    ├── README.md

    ├── clean.bat

    ├── clean.sh

    ├── CMakeLists.txt

    ├── install-deps

    ├── install-deps.bat

    ├── install-openblas.sh

    ├── install.bat

    ├── install.sh

    ├── LICENSE.md

    ├── prepare_mklml.sh

    ├── test.bat

    ├── test.sh

    ├── travis_cuda_install.sh

    ├── uninstall.bat

    ├── update.sh

    ├── .travis.yml

    ├── admin/

    │   └── updateSubmodules.sh

    ├── cmake/

    │   ├── FindTorch.cmake

    │   └── 3.6/

    │       ├── CMakeLists.txt

    │       └── Modules/

    │           ├── FindCUDA.cmake

    │           └── FindCUDA/

    │               ├── make2cmake.cmake

    │               ├── parse_cubin.cmake

    │               ├── run_nvcc.cmake

    │               └── select_compute_arch.cmake

    ├── exe/

    │   └── CMakeLists.txt

    ├── extra/

    │   └── CMakeLists.txt

    └── win-files/

        ├── README.md

        ├── patch/

        │   ├── lua-cjson.patch

        │   ├── luaffifb.patch

        │   └── torch.patch

        └── test/

            ├── ffifb_perf_ffi.lua

            └── ffifb_perf_purelua.lua

Files Content

================================================
FILE: README.md
================================================
DISCONTINUATION OF PROJECT.

This project will no longer be maintained by Intel.

Intel has ceased development and contributions including, but not limited to, maintenance, bug fixes, new releases, or updates, to this project. 

Intel no longer accepts patches to this project.

If you have an ongoing need to use this project, are interested in independently developing it, or would like to maintain patches for the open source software community, please create your own fork of this project. 
[![Build Status](https://travis-ci.org/torch/distro.svg?branch=master)](https://travis-ci.org/torch/distro)

Torch*
============
Torch is a scientific computing framework with wide support for machine learning algorithms. It is easy to use and efficient, thanks to an easy and fast scripting language, LuaJIT, and an underlying C/CUDA implementation.

Intel®Software Optimization for Torch
============
This fork is dedicated to improving Torch performance when running on CPU, in particular Intel® Xeon processors (HSW, BDW, Xeon Phi)


### Requirements
If you are root user, please use this command to install OpenBlas and other dependency: 
```sh
bash install-deps
```

If you are not root user, please use this command to install OpenBlas from source code, and this script will add OpenBLAS to LD_LIBRARY_PATH automaticly.
```sh
. ./install-openblas.sh
```

### Building
Install this repo, which installs the torch distribution, with a lot of nice goodies. 

You can specify which compiler to compile the project, and the default compiler is gcc & g++.
```sh
git clone https://github.com/intel/torch.git ~/torch
cd ~/torch; bash install-deps;
./install.sh        #use gcc to install torch
./install.sh icc  #use icc to install torch
```

By default Torch will install LuaJIT 2.1. If you want other options, you can use the command:
```sh
TORCH_LUA_VERSION=LUA51 ./install.sh
TORCH_LUA_VERSION=LUA52 ./install.sh
```


### Cleaning
To remove all the temporary compilation files you can run:
```sh
./clean.sh
```

### Test
You can test that all libraries are installed properly by running:
```sh
./test.sh
```
Tested on Ubuntu 14.04, CentOS 7.

### More build options for install.sh

```sh
./install.sh [gcc] [avx512] [mklml] [noskip]
```

*  icc/gcc,    default gcc
*  avx512/off, default off, avx512 will force compilers(GCC version should be greater than 4.9.2) to use AVX512F instructions to compile the framework. 
*  mklml/mkl,  default mkl.
*  noskip/skip,default noskip, skip means skip the openblas checking  

If you want to use MKL as the default BLAS library, please activate MKL before install.sh:
```sh
source /opt/intel/mkl/bin/mklvars.sh intel64
```

---
>\* Other names and trademarks may be claimed as the property of others.





================================================
FILE: clean.bat
================================================
@setlocal enableextensions
@echo off

::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
:::: This script cleans temporary compilation file          ::::
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::


set TORCH_DISTRO=%~dp0.

for /d %%G in ("%TORCH_DISTRO%\win-files\3rd\wineditline-*") do rmdir /s /q "%%~G"\build
cd %TORCH_DISTRO%\win-files\3rd\dlfcn-win32 && git clean -fdx

cd %TORCH_DISTRO%\exe\lua-5.1.5 && del /q *.obj *.o *.lib *.dll *.exp *.exe
cd %TORCH_DISTRO%\exe\lua-5.2.4 && del /q *.obj *.o *.lib *.dll *.exp *.exe
cd %TORCH_DISTRO%\exe\lua-5.3.3 && del /q *.obj *.o *.lib *.dll *.exp *.exe
cd %TORCH_DISTRO%\exe\luajit-2.0 && git clean -fdx
cd %TORCH_DISTRO%\exe\luajit-2.1 && git clean -fdx
cd %TORCH_DISTRO%\exe\luarocks && git clean -fdx

cd %TORCH_DISTRO% && git submodule foreach --recursive git clean -fdx

echo Cleaning is finished
@endlocal



================================================
FILE: clean.sh
================================================
#!/usr/bin/env bash

THIS_DIR=$(cd $(dirname $0); pwd)
find "${THIS_DIR}/" -type d -name build -o -name install \
    | grep -v '/exe/luajit-rocks/' | xargs -I {} rm -rf {}

rm -rf extra/threads/CMakeFiles
rm -rf extra/threads/build.luarocks
rm -rf extra/threads/CMakeCache.txt
rm -rf extra/threads/Makefile
rm -rf extra/threads/cmake_install.cmake
rm -rf extra/threads/install_manifest.txt

rm -rf extra/luafilesystem/lfs.so



================================================
FILE: CMakeLists.txt
================================================
cmake_minimum_required(VERSION 2.8)
ADD_SUBDIRECTORY(exe)

# If this variable is defined, a pure cmake build is done end-to-end with no luarocks
IF (WITH_LUAROCKS STREQUAL OFF)

  # LUA_INCDIR - place where lua headers exist
  # LUA_LIBDIR - place where lua libraries exist
  # LUADIR - LUA_PATH
  # LIBDIR - LUA_CPATH
  # LUALIB - the lua library to link against

  IF (NOT DEFINED ${LUA_INCDIR})
     SET(LUA_INCDIR ${CMAKE_INSTALL_PREFIX}/include)
     MESSAGE(STATUS "LUA_INCDIR: ${LUA_INCDIR}")
  ENDIF(NOT DEFINED ${LUA_INCDIR})
  IF (NOT DEFINED ${LUA_LIBDIR})
     SET(LUA_LIBDIR ${CMAKE_INSTALL_PREFIX}/lib)
     MESSAGE(STATUS "LUA_LIBDIR: ${LUA_LIBDIR}")
  ENDIF(NOT DEFINED ${LUA_LIBDIR})
  IF (NOT DEFINED ${LUA_BINDIR})
     SET(LUA_BINDIR ${CMAKE_INSTALL_PREFIX}/bin)
     MESSAGE(STATUS "LUA_BINDIR: ${LUA_BINDIR}")
  ENDIF(NOT DEFINED ${LUA_BINDIR})
  IF (NOT DEFINED ${LUALIB})
     IF (WITH_LUAJIT21 OR WITH_LUAJIT20)
       SET(LUALIB ${LIBRARY_OUTPUT_PATH}/libluajit.so)
     ELSE (WITH_LUAJIT21 OR WITH_LUAJIT20)
       SET(LUALIB ${LIBRARY_OUTPUT_PATH}/liblua.so)
     ENDIF (WITH_LUAJIT21 OR WITH_LUAJIT20)
     MESSAGE(STATUS "LUALIB: ${LUALIB}")
  ENDIF(NOT DEFINED ${LUALIB})
  IF (NOT DEFINED ${LUADIR})
     SET(LUADIR ${CMAKE_INSTALL_PREFIX}/${INSTALL_LUA_PATH_SUBDIR})
     MESSAGE(STATUS "LUADIR: ${LUADIR}")
  ENDIF(NOT DEFINED ${LUADIR})
  IF (NOT DEFINED ${LIBDIR})
     SET(LIBDIR ${CMAKE_INSTALL_PREFIX}/${INSTALL_LUA_CPATH_SUBDIR})
     MESSAGE(STATUS "LIBDIR: ${LIBDIR}")
  ENDIF(NOT DEFINED ${LIBDIR})
  IF (NOT DEFINED ${LUA_BINDIR})
    SET(LUA_BINDIR ${CMAKE_INSTALL_PREFIX}/bin)
    MESSAGE(STATUS "LUA_BINDIR: ${LUA_BINDIR}")
  ENDIF(NOT DEFINED ${LUA_BINDIR})
  IF (NOT DEFINED ${SCRIPTS_DIR})
    SET(SCRIPTS_DIR ${CMAKE_INSTALL_PREFIX}/bin)
    MESSAGE(STATUS "SCRIPTS_DIR: ${SCRIPTS_DIR}")
  ENDIF(NOT DEFINED ${SCRIPTS_DIR})

  INCLUDE_DIRECTORIES(${LUA_INCDIR})
  IF (WITH_LUAJIT21 OR WITH_LUAJIT20)
    INCLUDE_DIRECTORIES(${LuaJIT_SOURCE_DIR}/src ${LuaJIT_BINARY_DIR})
    SET(LUA "luajit")
  ELSE (WITH_LUAJIT21 OR WITH_LUAJIT20)
    INCLUDE_DIRECTORIES(${Lua_SOURCE_DIR}/src ${Lua_BINARY_DIR})
    SET(LUA "lua")
  ENDIF (WITH_LUAJIT21 OR WITH_LUAJIT20)

  # add torch defs
  # this is done to fake compliance with FIND_PACKAGE(Torch)
  SET(CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/cmake"
    "${CMAKE_CURRENT_BINARY_DIR}/cmake" "${CMAKE_MODULE_PATH}")

  INCLUDE_DIRECTORIES("${CMAKE_CURRENT_SOURCE_DIR}/pkg/torch/lib/TH")
  INCLUDE_DIRECTORIES("${CMAKE_CURRENT_BINARY_DIR}/pkg/torch/lib/TH")
  INCLUDE_DIRECTORIES("${CMAKE_CURRENT_SOURCE_DIR}/pkg/torch/lib/luaT")
  INCLUDE_DIRECTORIES("${CMAKE_CURRENT_BINARY_DIR}/pkg/torch/lib/luaT")

  SET(Torch_INSTALL_INCLUDE_SUBDIR ${LUA_INCDIR})
  SET(Torch_INSTALL_INCLUDE ${LUA_INCDIR})
  SET(Torch_INSTALL_LUA_PATH_SUBDIR ${LUADIR})
  SET(Torch_INSTALL_LUA_CPATH_SUBDIR ${LIBDIR})
  SET(Torch_INSTALL_LIB ${LUA_LIBDIR})
  SET(Torch_INSTALL_LIB_SUBDIR ${LUA_LIBDIR})

  # Finally add the two directories
  ADD_SUBDIRECTORY(pkg)
  ADD_SUBDIRECTORY(extra)

ENDIF (WITH_LUAROCKS STREQUAL OFF)



================================================
FILE: install-deps
================================================
#!/usr/bin/env bash
set -e

######################################################################
# This script installs required dependencies for Torch7
######################################################################
{

install_openblas() {
    # Get and build OpenBlas (Torch is much better with a decent Blas)
    # Optionally set environment variable PREFIX to control
    # the installation directory.


    local PATH=$PATH:/sbin ## improve chances of finding ldconfig
    # Only proceed installing OpenBLAS if either ldconfig is unavailable, or ldconfig
    # reports that OpenBLAS is not already installed.
    if ! type ldconfig >/dev/null || ! ldconfig -p | grep -q "lib\(open\)\?blas.so"; then
        local tempdir=$(mktemp -d)

        git clone https://github.com/xianyi/OpenBLAS.git "$tempdir"/OpenBLAS  || { echo "Error. Cannot clone OpenBLAS." >&2 ; exit 1 ; }
        cd "$tempdir"/OpenBLAS || { echo "Error. Cannot create tempdir." >&2 ; exit 1 ; }
        if [ $(getconf _NPROCESSORS_CONF) == 1 ]; then
            make NO_AFFINITY=1 USE_OPENMP=0 USE_THREAD=0
        else
            make NO_AFFINITY=1 USE_OPENMP=1
        fi
        RET=$?;
        if [ $RET -ne 0 ]; then
            echo "Error. OpenBLAS could not be compiled";
            exit $RET;
        fi
        if [ ! -z "$PREFIX" ]; then
            sudo make install PREFIX="$PREFIX"
        else
            sudo make install
        fi
        RET=$?;
        if [ $RET -ne 0 ]; then
            echo "Error. OpenBLAS could not be installed";
            exit $RET;
        fi
        cd -
        rm -rf "$tempdir"
    else
        echo "Skipping install of OpenBLAS - it is already installed." >&2
    fi
}

install_openblas_AUR() {
    # build and install an OpenBLAS package for Archlinux
    cd /tmp && \
    curl https://aur.archlinux.org/cgit/aur.git/snapshot/openblas-lapack.tar.gz | tar zxf - && \
    cd openblas-lapack
    makepkg -csi --noconfirm
    RET=$?;
    if [ $RET -ne 0 ]; then
        echo "Error. OpenBLAS could not be installed";
        exit $RET;
    fi
}

checkupdates_archlinux() {
    # checks if archlinux is up to date
    if [[ -n $(checkupdates) ]]; then
        echo "It seems that your system is not up to date."
        echo "It is recommended to update your system before going any further."
        read -p "Continue installation ? [y/N] " yn
            case $yn in
                Y|y ) echo "Continuing...";;
                * ) echo "Installation aborted."
                    echo "Relaunch this script after updating your system with 'pacman -Syu'."
                    exit 0
            esac
    fi
}

# Based on Platform:
if [[ `uname` == 'Darwin' ]]; then
    # GCC?
    if [[ `which gcc` == '' ]]; then
        echo "MacOS doesn't come with GCC: please install XCode and the command line tools."
        exit 1
    fi

    # Install Homebrew (pkg manager):
    if [[ `which brew` == '' ]]; then
        ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"
    fi

    # Install dependencies:
    brew update
    brew install git readline cmake wget
    brew install libjpeg imagemagick zeromq graphicsmagick openssl
    brew link readline --force
    brew cask install xquartz
    brew list -1 | grep -q "^gnuplot\$" && brew remove gnuplot
    brew install gnuplot --with-wxmac --with-cairo --with-pdflib-lite --with-x11 --without-lua
    brew install qt || true

elif [[ "$(uname)" == 'Linux' ]]; then

    if [[ -r /etc/os-release ]]; then
        # this will get the required information without dirtying any env state
        DIST_VERS="$( ( . /etc/os-release &>/dev/null
                        echo "$ID $VERSION_ID") )"
        DISTRO="${DIST_VERS%% *}" # get our distro name
        VERSION="${DIST_VERS##* }" # get our version number
    elif [[ -r /etc/redhat-release ]]; then
        DIST_VERS=( $( cat /etc/redhat-release ) ) # make the file an array
        DISTRO="${DIST_VERS[0],,}" # get the first element and get lcase
        VERSION="${DIST_VERS[2]}" # get the third element (version)
    elif [[ -r /etc/lsb-release ]]; then
        DIST_VERS="$( ( . /etc/lsb-release &>/dev/null
                        echo "${DISTRIB_ID,,} $DISTRIB_RELEASE") )"
        DISTRO="${DIST_VERS%% *}" # get our distro name
        VERSION="${DIST_VERS##* }" # get our version number
    else # well, I'm out of ideas for now
        echo '==> Failed to determine distro and version.'
        exit 1
    fi

    # Detect fedora
    if [[ "$DISTRO" = "fedora" ]]; then
        distribution="fedora"
        fedora_major_version="$VERSION"
    # Detect archlinux
    elif [[ "$DISTRO" = "arch" ]]; then
        distribution="archlinux"
    # Detect Ubuntu
    elif [[ "$DISTRO" = "ubuntu" ]]; then
        export DEBIAN_FRONTEND=noninteractive
        distribution="ubuntu"
        ubuntu_major_version="${VERSION%%.*}"
    elif [[ "$DISTRO" = "linuxmint" ]]; then
        export DEBIAN_FRONTEND=noninteractive
        distribution="ubuntu"
        ubuntu_major_version="16"
    # Detect elementary OS
    elif [[ "$DISTRO" = "elementary" ]]; then
        export DEBIAN_FRONTEND=noninteractive
        distribution="elementary"
        elementary_version="${VERSION%.*}"
    # Detect CentOS
    elif [[ "$DISTRO" = "centos" ]]; then
        distribution="centos"
        centos_major_version="$VERSION"
    # Detect AWS
    elif [[ "$DISTRO" = "amzn" ]]; then
        distribution="amzn"
        amzn_major_version="$VERSION"
    elif [[ "$DISTRO" == "raspbian" ]]; then
        distribution="raspbian"
        debian_major_version="$VERSION"
    elif [[ "$DISTRO" == "opensuse" ]]; then
	distribution="opensuse"
	opensuse_major_version="$VERSION"
    # Detect Debian
    elif [[ "$DISTRO" = "debian" ]]; then
	distribution="debian"
	debian_major_version="$VERSION"
    elif [[ "$DISTRO" = "neon" ]]; then
	distribution="ubuntu"
	ubuntu_major_version="${VERSION%%.*}"
    else
        echo '==> Only Ubuntu, elementary OS, Fedora, Archlinux, OpenSUSE, Debian, CentOS and KDE neon distributions are supported.'
        exit 1
    fi

    # Install dependencies for Torch:
    if [[ $distribution == 'ubuntu' ]]; then
        if sudo apt-get update ; then
            echo "Updated successfully."
        else
            echo "Some portion of the update is failed"
        fi
        # python-software-properties is required for apt-add-repository
        sudo apt-get install -y python-software-properties
        echo "==> Found Ubuntu version ${ubuntu_major_version}.xx"
        if [[ $ubuntu_major_version -lt '12' ]]; then
            echo '==> Ubuntu version not supported.'
            exit 1
        elif [[ $ubuntu_major_version -lt '14' ]]; then # 12.xx
            sudo -E add-apt-repository -y ppa:chris-lea/zeromq
        elif [[ $ubuntu_major_version -lt '15' ]]; then # 14.xx
	    sudo -E apt-get install -y software-properties-common
            sudo -E add-apt-repository -y ppa:jtaylor/ipython
        else
            sudo apt-get install -y software-properties-common \
                libgraphicsmagick1-dev libfftw3-dev sox libsox-dev \
                libsox-fmt-all
        fi

        if sudo apt-get update ; then
            echo "Updated successfully."
        else
            echo "Some portion of the update is failed"
        fi
        sudo apt-get install -y build-essential gcc g++ curl \
            cmake libreadline-dev git-core libqt4-dev libjpeg-dev \
            libpng-dev ncurses-dev imagemagick libzmq3-dev gfortran \
            unzip gnuplot gnuplot-x11 ipython

        gcc_major_version=$(gcc --version | grep ^gcc | awk '{print $4}' | \
                            cut -c 1)
        if [[ $gcc_major_version == '5' ]]; then
            echo '==> Found GCC 5, installing GCC 4.9.'
            sudo apt-get install -y gcc-4.9 libgfortran-4.9-dev g++-4.9
        fi

        if [[ $ubuntu_major_version -lt '15' ]]; then
            sudo apt-get install libqt4-core libqt4-gui
        fi

        install_openblas || true

    elif [[ $distribution == 'raspbian' ]]; then
        echo "==> Found Raspbian version ${debian_major_version}.xx"
        if sudo apt-get update ; then
            echo "Updated successfully."
        else
            echo "Some portion of the update is failed"
        fi
        sudo apt-get install -y build-essential gcc g++ curl \
            cmake libreadline-dev git-core libqt4-dev libjpeg-dev \
            libpng-dev ncurses-dev imagemagick libzmq3-dev gfortran \
            unzip gnuplot gnuplot-x11 ipython

        install_openblas || true

    elif [[ $distribution == 'opensuse' ]]; then
        if sudo zypper refresh ; then
            echo "Updated successfully."
        else
            echo "Some portion of the update failed"
        fi
        sudo zypper install -y -t pattern devel_basis
	if [[ $opensuse_major_version == '42.2' ]]; then
	    sudo zypper install -y gcc curl cmake readline-devel IPython git-core \
            	libqt4 libjpeg62-devel libpng16-compat-devel ImageMagick unzip \
            	gnuplot gcc-fortran libzmq5
	else
            sudo zypper install -y gcc curl cmake readline-devel IPython git-core \
            	libqt4 libjpeg8-devel libpng15-compat-devel imagemagick unzip \
            	gnuplot gcc-fortran libzmq3
	fi

    elif [[ $distribution == 'elementary' ]]; then
        declare -a target_pkgs
        target_pkgs=( build-essential gcc g++ curl \
                      cmake libreadline-dev git-core libqtcore4 libqtgui4 \
                      libqt4-dev libjpeg-dev libpng-dev ncurses-dev \
                      imagemagick libzmq3-dev gfortran unzip gnuplot \
                      gnuplot-x11 ipython )
        if sudo apt-get update ; then
            echo "Updated successfully."
        else
            echo "Some portion of the update is failed"
        fi
        # python-software-properties is required for apt-add-repository
        sudo apt-get install -y python-software-properties
        if [[ $elementary_version == '0.3' ]]; then
            echo '==> Found Ubuntu version 14.xx based elementary installation, installing dependencies'
            sudo apt-get install -y software-properties-common \
                libgraphicsmagick1-dev libfftw3-dev sox libsox-dev \
                libsox-fmt-all

            sudo -E add-apt-repository -y ppa:jtaylor/ipython
        else
            sudo -E add-apt-repository -y ppa:chris-lea/zeromq
        fi
        if sudo apt-get update ; then
            echo "Updated successfully."
        else
            echo "Some portion of the update is failed"
        fi
        sudo apt-get install -y "${target_pkgs[@]}"

        install_openblas || true

    elif [[ $distribution == 'archlinux' ]]; then
        echo "Archlinux installation"
        checkupdates_archlinux
        sudo pacman -S --quiet --noconfirm --needed \
            cmake curl readline ncurses git \
            gnuplot unzip libjpeg-turbo libpng libpng \
            imagemagick graphicsmagick fftw sox zeromq \
            ipython qt4 qt5-webkit-ng || exit 1
        pacman -Sl multilib &>/dev/null
        if [[ $? -ne 0 ]]; then
            gcc_package="gcc"
        else
            gcc_package="gcc-multilib"
        fi
        sudo pacman -S --quiet --noconfirm --needed \
            ${gcc_package} gcc-fortran || exit 1
        # if openblas is not installed yet
        pacman -Qs openblas &> /dev/null
        if [[ $? -ne 0 ]]; then
            install_openblas_AUR || true
        else
            echo "OpenBLAS is already installed"
        fi

    elif [[ $distribution == 'fedora' ]]; then
        if [[ $fedora_major_version == '20' ]]; then
            sudo yum install -y cmake curl readline-devel ncurses-devel \
                                gcc-c++ gcc-gfortran git gnuplot unzip \
                                libjpeg-turbo-devel libpng-devel \
                                ImageMagick GraphicsMagick-devel fftw-devel \
                                sox-devel sox zeromq3-devel \
                                qt-devel qtwebkit-devel sox-plugins-freeworld \
                                ipython
            install_openblas || true
        elif [[ $fedora_major_version == '22' ||  $fedora_major_version == '23' ]]; then
            #using dnf - since yum has been deprecated
            #sox-plugins-freeworld is not yet available in repos for F22
            sudo dnf install -y make cmake curl readline-devel ncurses-devel \
                                gcc-c++ gcc-gfortran git gnuplot unzip \
                                libjpeg-turbo-devel libpng-devel \
                                ImageMagick GraphicsMagick-devel fftw-devel \
                                sox-devel sox qt-devel qtwebkit-devel \
                                python-ipython czmq czmq-devel
            install_openblas || true
        elif [[ $fedora_major_version == '24' ||  $fedora_major_version == '25' ]]; then
            sudo dnf install -y make cmake curl readline-devel ncurses-devel \
                                gcc-c++ gcc-gfortran git gnuplot unzip \
                                libjpeg-turbo-devel libpng-devel \
                                ImageMagick GraphicsMagick-devel fftw-devel \
                                sox-devel sox qt-devel qtwebkit-devel \
                                python-ipython czmq czmq-devel
            install_openblas || true
        else
            echo "Only Fedora 20-25 are supported for now, aborting."
            exit 1
        fi
    elif [[ $distribution == 'centos' ]]; then
        if [[ $centos_major_version == '7' ]]; then
            sudo yum install -y epel-release # a lot of things live in EPEL
            sudo yum install -y make cmake curl readline-devel ncurses-devel \
                                gcc-c++ gcc-gfortran git gnuplot unzip \
                                libjpeg-turbo-devel libpng-devel \
                                ImageMagick GraphicsMagick-devel fftw-devel \
                                sox-devel sox zeromq3-devel \
                                qt-devel qtwebkit-devel sox-plugins-freeworld
            sudo yum install -y python-ipython
            install_openblas || true
        else
            echo "Only CentOS 7 is supported for now, aborting."
            exit 1
        fi
    elif [[ $distribution == 'amzn' ]]; then
        sudo yum install -y cmake curl readline-devel ncurses-devel \
                            gcc-c++ gcc-gfortran git gnuplot unzip \
                            libjpeg-turbo-devel libpng-devel \
                            ImageMagick GraphicsMagick-devel fftw-devel \
                            libgfortran python27-pip git openssl-devel

        #
        # These libraries are missing from amzn linux
        # sox-devel sox sox-plugins-freeworld qt-devel qtwebkit-devel
        #

        sudo yum --enablerepo=epel install -y zeromq3-devel
        sudo pip install ipython

        install_openblas || true
    elif [[ $distribution == 'debian' ]]; then
	if [[ $debian_major_version == '8' ]] || [[ $debian_major_version == '9' ]]; then
	    echo "==> Found Debian version ${debian_major_version}"
	    if sudo apt-get update ; then
	       echo "Updated successfully."
	    else
	       echo "Some portion of the update is failed"
	    fi
	    #Basic package required for Torch
	    sudo apt-get install -y build-essential curl \
		 cmake libreadline-dev git-core libqt4-dev libjpeg-dev \
		 libpng-dev ncurses-dev imagemagick libzmq3-dev gfortran \
		 unzip gnuplot gnuplot-x11 ipython

	    #require for common torch plug-ins
	    sudo apt-get install -y libgraphicsmagick1-dev libfftw3-dev sox libsox-dev

	    install_openblas || true
	else
	    echo "Only Jessie Debian 8 and 9 is supported for now, aborting."
	    exit 1
	fi
    fi
elif [[ "$(uname)" == 'FreeBSD' ]]; then
    pkg install ImageMagick cmake curl fftw3 git gnuplot libjpeg-turbo \
        libzmq3 ncurses openblas openssl png py27-ipython \
        py27-pip qt4-corelib qt4-gui readline unzip

else
    # Unsupported
    echo '==> platform not supported, aborting'
    exit 1
fi

ipython_exists=$(command -v ipython) || true
if [[ $ipython_exists ]]; then {
    ipython_version=$(ipython --version|cut -f1 -d'.')
    if [[ $ipython_version -lt 2 ]]; then {
        echo 'WARNING: Your ipython version is too old.  Type "ipython --version" to see this.  Should be at least version 2'
    } fi
} fi

# Done.
echo "==> Torch7's dependencies have been installed"

}



================================================
FILE: install-deps.bat
================================================
@echo off

::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
:::: This script setup directories, dependencies for Torch7 ::::
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::


:::: Customizable variables ::::


:: which lua version will be installed for Torch7, default to LUAJIT21
:: accepted lua versions: LUAJIT21, LUAJIT20, LUA53, LUA52, LUA51
REM  set TORCH_LUA_VERSION=LUAJIT21

:: where to install Torch7, default to install\ under distro\
REM  set TORCH_INSTALL_DIR=D:\Torch

:: conda environment name for Torch7, default to torch-vcversion
REM  set TORCH_CONDA_ENV=mytorch7

:: which blas/lapack libraries will be used, default to openblas installed by conda
:: [1] mkl: download from https://software.intel.com/intel-mkl, install and set following two variables
REM  set INTEL_MKL_DIR=D:\\Intel\\SWTools\\compilers_and_libraries\\windows\\mkl\\
REM  set INTEL_COMPILER_DIR=D:\\Intel\\SWTools\\compilers_and_libraries\\windows\\compiler\\
:: [2] other: set path to the blas library and path to the laback library
:: both BLAS and LAPACK should be set even if they refer to the same library
:: take openblas for example: download latest release from https://github.com/xianyi/OpenBLAS/releases/latest
:: use mingw cross compiler tools in cygwin, since mingw windows native gfortrain is available in cygwin but not in msys2
:: compilation command in cygwin: make CC=x86_64-w64-mingw32-gcc FC=x86_64-w64-mingw32-gfortran CROSS_SUFFIX=x86_64-w64-mingw32-
:: please refer to openblas's README for detailed installation instructions
REM  set BLAS_LIBRARIES=D:\\Libraries\\lib\libopenblas.dll.a
REM  set LAPACK_LIBRARIES=D:\\Libraries\\lib\libopenblas.dll.a

:: where to find cudnn library
REM  set CUDNN_PATH=D:\NVIDIA\CUDNN\v5.1\bin\cudnn64_5.dll

:: whether update dependencies if already setup, default to not update
REM  set TORCH_UPDATE_DEPS=

::::  End of customization  ::::


set ECHO_PREFIX=+++++++
set TORCH_SETUP_FAIL=1

:::: validate msvc version  ::::

if "%VisualStudioVersion%" == "" (
  if not "%VS150COMNTOOLS%" == "" ( call "%VS150COMNTOOLS%..\..\VC\Auxiliary\Build\vcvarsall.bat" x64 && goto :VS_SETUP)
  if not "%VS140COMNTOOLS%" == "" ( call "%VS140COMNTOOLS%..\..\VC\vcvarsall.bat" x64 && goto :VS_SETUP)
  if not "%VS120COMNTOOLS%" == "" ( call "%VS120COMNTOOLS%..\..\VC\vcvarsall.bat" x64 && goto :VS_SETUP)
  if not "%VS110COMNTOOLS%" == "" ( call "%VS110COMNTOOLS%..\..\VC\vcvarsall.bat" x64 && goto :VS_SETUP)
  if not "%VS100COMNTOOLS%" == "" ( call "%VS100COMNTOOLS%..\..\VC\vcvarsall.bat" x64 && goto :VS_SETUP)
  if not "%VS90COMNTOOLS%"  == "" ( call "%VS90COMNTOOLS%..\..\VC\vcvarsall.bat"  x64 && goto :VS_SETUP)
)
:VS_SETUP

if "%VisualStudioVersion%" == "" (
  echo %ECHO_PREFIX% Can not find environment variable VisualStudioVersion, msvc is not setup porperly
  goto :FAIL
)

set TORCH_VS_VERSION=%VisualStudioVersion:.0=%

if "%PreferredToolArchitecture%" == "x64" (
  if "%CommandPromptType%" == "Cross" (
    if "%Platform%" == "ARM" set TORCH_VS_PLATFORM=amd64_arm
    if "%Platform%" == "X86" set TORCH_VS_PLATFORM=amd64_x86
    if "%Platform%" == "x86" set TORCH_VS_PLATFORM=amd64_x86
  )
) else (
  if "%CommandPromptType%" == "Cross" (
    if "%Platform%" == "ARM" set TORCH_VS_PLATFORM=x86_arm
    if "%Platform%" == "X64" set TORCH_VS_PLATFORM=x86_amd64
    if "%Platform%" == "x64" set TORCH_VS_PLATFORM=x86_amd64
  )
  if "%CommandPromptType%" == "Native" (
    if "%Platform%" == "X64" set TORCH_VS_PLATFORM=x64
    if "%Platform%" == "x64" set TORCH_VS_PLATFORM=x64
  )
  if "%Platform%"   == ""    set TORCH_VS_PLATFORM=x86
)

if     "%TORCH_VS_PLATFORM%" == "x86"                       set TORCH_VS_TARGET=x86
if not "%TORCH_VS_PLATFORM%" == "%TORCH_VS_PLATFORM:_x86=%" set TORCH_VS_TARGET=x86
if not "%TORCH_VS_PLATFORM%" == "%TORCH_VS_PLATFORM:_arm=%" set TORCH_VS_TARGET=arm
if     "%TORCH_VS_TARGET%"   == ""                          set TORCH_VS_TARGET=x64

::::  validate lua version  ::::

:: [TODO] currently luajit lua luarocks are installed from source. they can be changed to use luajit-rocks when
::        luajit-rocks is ready for windows

if "%TORCH_LUA_VERSION%" == "" set TORCH_LUA_VERSION=LUAJIT21
if /i "%TORCH_LUA_VERSION%" == "LUAJIT21" (
  set TORCH_LUAJIT_BRANCH=v2.1
  set TORCH_LUA_SOURCE=luajit-2.1
  set TORCH_LUAROCKS_LUA=5.1
)
if /i "%TORCH_LUA_VERSION%" == "LUAJIT20" (
  set TORCH_LUAJIT_BRANCH=master
  set TORCH_LUA_SOURCE=luajit-2.0
  set TORCH_LUAROCKS_LUA=5.1
)
if /i "%TORCH_LUA_VERSION%" == "LUA53" (
  set TORCH_LUA_SOURCE=lua-5.3.3
  set TORCH_LUAROCKS_LUA=5.3
)
if /i "%TORCH_LUA_VERSION%" == "LUA52" (
  set TORCH_LUA_SOURCE=lua-5.2.4
  set TORCH_LUAROCKS_LUA=5.2
)
if /i "%TORCH_LUA_VERSION%" == "LUA51" (
  set TORCH_LUA_SOURCE=lua-5.1.5
  set TORCH_LUAROCKS_LUA=5.1
)
if /i "%TORCH_LUA_SOURCE%" == "" (
  echo %ECHO_PREFIX% Bad lua version: %TORCH_LUA_VERSION%, only support LUAJIT21, LUAJIT20, LUA53, LUA52, LUA51
  goto :FAIL
)

::::    Setup directories   ::::

set TORCH_DISTRO=%~dp0.
if "%TORCH_INSTALL_DIR%" == "" set TORCH_INSTALL_DIR=%TORCH_DISTRO%\install
set TORCH_INSTALL_BIN=%TORCH_INSTALL_DIR%\bin
set TORCH_INSTALL_LIB=%TORCH_INSTALL_DIR%\lib
set TORCH_INSTALL_INC=%TORCH_INSTALL_DIR%\include
set TORCH_INSTALL_ROC=%TORCH_INSTALL_DIR%\luarocks
if not exist %TORCH_INSTALL_BIN% md %TORCH_INSTALL_BIN%
if not exist %TORCH_INSTALL_LIB% md %TORCH_INSTALL_LIB%
if not exist %TORCH_INSTALL_INC% md %TORCH_INSTALL_INC%
if not "%TORCH_LUAJIT_BRANCH%" == "" if not exist %TORCH_INSTALL_BIN%\lua\jit md %TORCH_INSTALL_BIN%\lua\jit
if not exist %TORCH_DISTRO%\win-files\3rd md %TORCH_DISTRO%\win-files\3rd

echo %ECHO_PREFIX% Torch7 will be installed under %TORCH_INSTALL_DIR% with %TORCH_LUA_SOURCE%, vs%TORCH_VS_VERSION% %TORCH_VS_PLATFORM%
echo %ECHO_PREFIX% Bin: %TORCH_INSTALL_BIN%
echo %ECHO_PREFIX% Lib: %TORCH_INSTALL_LIB%
echo %ECHO_PREFIX% Inc: %TORCH_INSTALL_INC%

::::   Setup dependencies   ::::

:: has blas/lapack?
if not "%INTEL_MKL_DIR%" == "" if exist %INTEL_MKL_DIR% ( set "TORCH_SETUP_HAS_MKL=1" && set "TORCH_SETUP_HAS_BLAS=1" && set "TORCH_SETUP_HAS_LAPACK=1" )
if not "%BLAS_LIBRARIES%" == "" if exist %BLAS_LIBRARIES% set TORCH_SETUP_HAS_BLAS=1
if not "%LAPACK_LIBRARIES%" == "" if exist %LAPACK_LIBRARIES% set TORCH_SETUP_HAS_LAPACK=1

:: has cuda?
for /f "delims=" %%i in ('where nvcc') do (
  set NVCC_CMD=%%i
  goto :AFTER_NVCC
)
:AFTER_NVCC
if not "%NVCC_CMD%" == "" set TORCH_SETUP_HAS_CUDA=1

:: has conda?
for /f "delims=" %%i in ('where conda') do (
  set CONDA_CMD=%%i
  goto :AFTER_CONDA
)
:AFTER_CONDA

if "%CONDA_CMD%" == "" (
  echo %ECHO_PREFIX% Can not find conda, some dependencies can not be resolved
  if not "%TORCH_SETUP_HAS_BLAS%" == "1" (
    echo %ECHO_PREFIX% Can not install torch, please either specify the blas library or install conda
    goto :FAIL
  )
  goto :NO_CONDA
)

set TORCH_CONDA_INFO=%TORCH_DISTRO%\win-files\check_conda_info_for_torch.txt
conda info > %TORCH_CONDA_INFO%
if "%TORCH_VS_TARGET%" == "x64" set TORCH_CONDA_PLATFORM=win-64
if "%TORCH_VS_TARGET%" == "arm" set TORCH_CONDA_PLATFORM=win-64
if "%TORCH_VS_TARGET%" == "x86" set TORCH_CONDA_PLATFORM=win-32

findstr "%TORCH_CONDA_PLATFORM%" "%TORCH_CONDA_INFO%" >nul
if errorlevel 1 (
  echo %ECHO_PREFIX% %TORCH_VS_TARGET% Torch7 requires %TORCH_CONDA_PLATFORM% conda, installation will continue without conda
  goto :NO_CONDA
)

if %TORCH_VS_VERSION% GEQ 14 ( set CONDA_VS_VERSION=14&& goto :CONDA_SETUP )
if %TORCH_VS_VERSION% GEQ 10 ( set CONDA_VS_VERSION=10&& goto :CONDA_SETUP )
set CONDA_VS_VERSION=9

:CONDA_SETUP

if "%TORCH_CONDA_ENV%" == "" set TORCH_CONDA_ENV=torch-vc%CONDA_VS_VERSION%

echo %ECHO_PREFIX% Createing conda environment '%TORCH_CONDA_ENV%' for Torch7 dependencies
conda create -n %TORCH_CONDA_ENV% -c conda-forge vc=%CONDA_VS_VERSION% --yes

set CONDA_DIR=%CONDA_CMD:\Scripts\conda.exe=%
set TORCH_CONDA_LIBRARY=%CONDA_DIR%\envs\%TORCH_CONDA_ENV%\Library
set TORCH_CONDA_LIBRARY=%TORCH_CONDA_LIBRARY:\=\\%
set PATH=%TORCH_CONDA_LIBRARY%\bin;%PATH%;
set NEW_PATH=%%CONDA_DIR%%\Scripts;%%CONDA_DIR%%\envs\%TORCH_CONDA_ENV%\Library\bin;%NEW_PATH%

set TORCH_CONDA_PKGS=%TORCH_DISTRO%\win-files\check_conda_packages_for_torch.txt
conda list -n %TORCH_CONDA_ENV% > %TORCH_CONDA_PKGS%

:: has cmake?
:: cmake should be installed before qt since its on qt5 while qtlua is on qt4
for /f "delims=" %%i in ('where cmake') do (
  set CMAKE_CMD=%%i
  goto :AFTER_CMAKE
)
if "%CMAKE_CMD%" == "" (
  echo %ECHO_PREFIX% Installing cmake by conda
  conda install -n %TORCH_CONDA_ENV% -c conda-forge cmake --yes
)
:AFTER_CMAKE

:: need openblas?
findstr "openblas" "%TORCH_CONDA_PKGS%" >nul
if not errorlevel 1 ( set "TORCH_SETUP_HAS_BLAS=1" && set "TORCH_SETUP_HAS_LAPACK=1" )

if not "%TORCH_SETUP_HAS_BLAS%"   == "1" goto :CONDA_INSTALL_OPENBLAS
if not "%TORCH_SETUP_HAS_LAPACK%" == "1" goto :CONDA_INSTALL_OPENBLAS
goto :AFTER_OPENBLAS

:CONDA_INSTALL_OPENBLAS
echo %ECHO_PREFIX% Installing openblas by conda, since there is no blas library specified
if not "%TORCH_VS_TARGET%" == "x86" conda install -n %TORCH_CONDA_ENV% -c ukoethe openblas --yes || goto :Fail
if     "%TORCH_VS_TARGET%" == "x86" conda install -n %TORCH_CONDA_ENV% -c omnia   openblas --yes || goto :Fail

:AFTER_OPENBLAS
if not "%TORCH_VS_TARGET%" == "x86" (
  if "%BLAS_LIBRARIES%"   == "" set BLAS_LIBRARIES=%TORCH_CONDA_LIBRARY%\\lib\\libopenblas.lib
  if "%LAPACK_LIBRARIES%" == "" set LAPACK_LIBRARIES=%TORCH_CONDA_LIBRARY%\\lib\\libopenblas.lib
)
if     "%TORCH_VS_TARGET%" == "x86" (
  if "%BLAS_LIBRARIES%"   == "" set BLAS_LIBRARIES=%TORCH_CONDA_LIBRARY%\\lib\\libopenblaspy.dll.a
  if "%LAPACK_LIBRARIES%" == "" set LAPACK_LIBRARIES=%TORCH_CONDA_LIBRARY%\\lib\\libopenblaspy.dll.a
)

:: other dependencies
findstr "jpeg" "%TORCH_CONDA_PKGS%" >nul
if errorlevel 1 set TORCH_DEPENDENCIES=%TORCH_DEPENDENCIES% jpeg
findstr "libpng" "%TORCH_CONDA_PKGS%" >nul
if errorlevel 1 set TORCH_DEPENDENCIES=%TORCH_DEPENDENCIES% libpng
findstr "zlib" "%TORCH_CONDA_PKGS%" >nul
if errorlevel 1 set TORCH_DEPENDENCIES=%TORCH_DEPENDENCIES% zlib
findstr "libxml2" "%TORCH_CONDA_PKGS%" >nul
if errorlevel 1 set TORCH_DEPENDENCIES=%TORCH_DEPENDENCIES% libxml2
findstr "qt" "%TORCH_CONDA_PKGS%" >nul
if errorlevel 1 set TORCH_DEPENDENCIES=%TORCH_DEPENDENCIES% qt=4.8.7

if not "%TORCH_DEPENDENCIES%" == "" (
  echo %ECHO_PREFIX% Installing %TORCH_DEPENDENCIES% by conda for Torch7
  conda install -n %TORCH_CONDA_ENV% -c conda-forge %TORCH_DEPENDENCIES% vc=%CONDA_VS_VERSION% --yes
)

:NO_CONDA
if exist "%TORCH_CONDA_INFO%" del /q %TORCH_CONDA_INFO%
if exist "%TORCH_CONDA_PKGS%" del /q %TORCH_CONDA_PKGS%

::::  git clone luarocks   ::::

echo %ECHO_PREFIX% Git clone luarocks for its tools
cd %TORCH_DISTRO%\exe\
if not exist luarocks\.git git clone https://github.com/keplerproject/luarocks.git luarocks
set PATH=%TORCH_DISTRO%\exe\luarocks\win32\tools\;%PATH%;

::::     install lua       ::::

echo %ECHO_PREFIX% Installing %TORCH_LUA_SOURCE%
cd %TORCH_DISTRO%\exe\
if not "%TORCH_LUAJIT_BRANCH%" == "" (
  if not exist %TORCH_LUA_SOURCE%\.git git clone -b %TORCH_LUAJIT_BRANCH% http://luajit.org/git/luajit-2.0.git %TORCH_LUA_SOURCE% || goto :Fail
  cd %TORCH_LUA_SOURCE% && ( if "%TORCH_UPDATE_DEPS%" == "1" git pull ) & cd src
) else (
  wget -nc https://www.lua.org/ftp/%TORCH_LUA_SOURCE%.tar.gz --no-check-certificate || goto :Fail
  7z x %TORCH_LUA_SOURCE%.tar.gz -y >NUL && 7z x %TORCH_LUA_SOURCE%.tar -y >NUL && cd %TORCH_LUA_SOURCE%\src
)
if not "%TORCH_LUAJIT_BRANCH%"=="" (
  call msvcbuild.bat || goto :FAIL
  copy /y jit\* %TORCH_INSTALL_BIN%\lua\jit\
  copy /y luajit.h %TORCH_INSTALL_INC%\luajit.h
  set LUAJIT_CMD=%TORCH_INSTALL_DIR%\luajit.cmd
) else (
  del /q *.obj *.o *.lib *.dll *.exp *.exe
  cl /nologo /c /O2 /W3 /D_CRT_SECURE_NO_DEPRECATE /MD /DLUA_BUILD_AS_DLL *.c || goto :FAIL
  ren lua.obj lua.o || goto :FAIL
  ren luac.obj luac.o || goto :FAIL
  link /nologo /DLL /IMPLIB:%TORCH_LUA_VERSION%.lib /OUT:%TORCH_LUA_VERSION%.dll *.obj || goto :FAIL
  link /nologo /OUT:lua.exe lua.o %TORCH_LUA_VERSION%.lib || goto :FAIL
  lib  /nologo /OUT:%TORCH_LUA_VERSION%-static.lib *.obj || goto :FAIL
  link /nologo /OUT:luac.exe luac.o %TORCH_LUA_VERSION%-static.lib || goto :FAIL
  copy /y lua.hpp %TORCH_INSTALL_INC%\lua.hpp
  set LUA_CMD=%TORCH_INSTALL_DIR%\lua.cmd
  set LUAC_CMD=%TORCH_INSTALL_DIR%\luac.cmd
)
copy /y *.exe %TORCH_INSTALL_BIN%\
copy /y *.dll %TORCH_INSTALL_BIN%\
copy /y *.lib %TORCH_INSTALL_LIB%\
for %%g in (lua.h,luaconf.h,lualib.h,lauxlib.h) do copy /y %%g %TORCH_INSTALL_INC%\%%g

::::   install luarocks    ::::

echo %ECHO_PREFIX% Installing luarocks
cd %TORCH_DISTRO%\exe\
if not exist luarocks\.git git clone https://github.com/keplerproject/luarocks.git luarocks
cd luarocks && ( if "%TORCH_UPDATE_DEPS%" == "1" git pull ) & call install.bat /F /Q /P %TORCH_INSTALL_ROC% /SELFCONTAINED /FORCECONFIG /NOREG /NOADMIN /LUA %TORCH_INSTALL_DIR% || goto :FAIL
for /f %%a in ('dir %TORCH_INSTALL_ROC%\config*.lua /b') do set LUAROCKS_CONFIG=%%a
set LUAROCKS_CONFIG=%TORCH_INSTALL_ROC%\%LUAROCKS_CONFIG%
echo rocks_servers = { >> %LUAROCKS_CONFIG%
echo   [[https://raw.githubusercontent.com/torch/rocks/master]], >> %LUAROCKS_CONFIG%
echo   [[https://raw.githubusercontent.com/rocks-moonscript-org/moonrocks-mirror/master]] >> %LUAROCKS_CONFIG%
echo } >> %LUAROCKS_CONFIG%

set LUAROCKS_CMD=%TORCH_INSTALL_DIR%\luarocks.cmd

:::: install wineditline   ::::

echo %ECHO_PREFIX% Installing wineditline for trepl package
cd %TORCH_DISTRO%\win-files\3rd\
wget -nc https://sourceforge.net/projects/mingweditline/files/wineditline-2.201.zip/download --no-check-certificate -O wineditline.zip
7z x wineditline.zip -y >NUL
cd wineditline*
cmake -E make_directory build && cd build && cmake .. -G "NMake Makefiles" -DLIB_SUFFIX="64" -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=..\ && nmake install

::::  install dlfcn-win32  ::::

echo %ECHO_PREFIX% Installing dlfcn-win32 for thread package
cd %TORCH_DISTRO%\win-files\3rd\
if not exist dlfcn-win32\.git git clone https://github.com/dlfcn-win32/dlfcn-win32.git
cd dlfcn-win32 && ( if "%TORCH_UPDATE_DEPS%" == "1" git pull )
cmake -E make_directory build && cd build && cmake .. -G "NMake Makefiles" -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=..\ && nmake install
set WIN_DLFCN_INCDIR=%TORCH_DISTRO:\=\\%\\win-files\\3rd\\dlfcn-win32\\include
set WIN_DLFCN_LIBDIR=%TORCH_DISTRO:\=\\%\\win-files\\3rd\\dlfcn-win32\\lib
copy /y %TORCH_DISTRO%\win-files\3rd\dlfcn-win32\bin\*.dll %TORCH_INSTALL_BIN%\

::::   download graphviz   ::::

echo %ECHO_PREFIX% Downloading graphviz for graph package
cd %TORCH_DISTRO%\win-files\3rd\
wget -nc https://github.com/mahkoCosmo/GraphViz_x64/raw/master/graphviz-2.38_x64.tar.gz --no-check-certificate -O graphviz-2.38_x64.tar.gz
7z x graphviz-2.38_x64.tar.gz -y && 7z x graphviz-2.38_x64.tar -ographviz-2.38_x64 -y >NUL
if not exist %TORCH_INSTALL_BIN%\graphviz md %TORCH_INSTALL_BIN%\graphviz
copy /y %TORCH_DISTRO%\win-files\3rd\graphviz-2.38_x64\bin\ %TORCH_INSTALL_BIN%\graphviz\

set NEW_PATH=%NEW_PATH%;%%TORCH_INSTALL_DIR%%\bin\graphviz

::::    create cmd utils   ::::

if not "%LUAJIT_CMD%" == "" (
  echo %ECHO_PREFIX% Creating torch-activate.cmd luajit.cmd luarocks.cmd cmake.cmd
) else (
  echo %ECHO_PREFIX% Creating torch-activate.cmd lua.cmd luac.cmd luarocks.cmd cmake.cmd
)

set NEW_PATH=%%TORCH_INSTALL_DIR%%;%%TORCH_INSTALL_DIR%%\bin;%%TORCH_INSTALL_DIR%%\luarocks;%%TORCH_INSTALL_DIR%%\luarocks\tools;%%TORCH_INSTALL_DIR%%\luarocks\systree\bin;%NEW_PATH%;%%PATH%%;;
set NEW_LUA_PATH=%%TORCH_INSTALL_DIR%%\luarocks\lua\?.lua;%%TORCH_INSTALL_DIR%%\luarocks\lua\?\init.lua;%%TORCH_INSTALL_DIR%%\luarocks\systree\share\lua\%TORCH_LUAROCKS_LUA%\?.lua;%%TORCH_INSTALL_DIR%%\luarocks\systree\share\lua\%TORCH_LUAROCKS_LUA%\?\init.lua;;
set NEW_LUA_CPATH=%%TORCH_INSTALL_DIR%%\luarocks\systree\lib\lua\%TORCH_LUAROCKS_LUA%\?.dll;;

set TORCHACTIVATE_CMD=%TORCH_INSTALL_DIR%\torch-activate.cmd
if exist %TORCHACTIVATE_CMD% del %TORCHACTIVATE_CMD%
echo @echo off>> %TORCHACTIVATE_CMD%
echo set CONDA_DIR=%CONDA_DIR%>> %TORCHACTIVATE_CMD%
echo set TORCH_INSTALL_DIR=%%~dp0.>> %TORCHACTIVATE_CMD%
echo set TORCH_CONDA_ENV=%TORCH_CONDA_ENV%>> %TORCHACTIVATE_CMD%
echo set TORCH_VS_VERSION=%TORCH_VS_VERSION%>> %TORCHACTIVATE_CMD%
echo set TORCH_VS_PLATFORM=%TORCH_VS_PLATFORM%>> %TORCHACTIVATE_CMD%
if "%TORCH_VS_VERSION%" == "15" (
  set VCVARSALL_BAT_PATH=..\..\VC\Auxiliary\Build\vcvarsall.bat
) else (
  set VCVARSALL_BAT_PATH=..\..\VC\vcvarsall.bat
)
echo for /f "delims=" %%%%i in ('call echo %%%%VS%TORCH_VS_VERSION%0COMNTOOLS%%%%') do call "%%%%i%VCVARSALL_BAT_PATH%" %TORCH_VS_PLATFORM%>> %TORCHACTIVATE_CMD%
echo set PATH=%NEW_PATH%>> %TORCHACTIVATE_CMD%
echo set LUA_PATH=%NEW_LUA_PATH%>> %TORCHACTIVATE_CMD%
echo set LUA_CPATH=%NEW_LUA_CPATH%>> %TORCHACTIVATE_CMD%
if not "%CUDNN_PATH%" == "" echo set CUDNN_PATH=%CUDNN_PATH%>> %TORCHACTIVATE_CMD%

if not "%LUAJIT_CMD%" == "" (
  if exist "%LUAJIT_CMD%" del %LUAJIT_CMD%
  echo @echo off>> "%LUAJIT_CMD%"
  echo setlocal>> "%LUAJIT_CMD%"
  echo set TORCH_INSTALL_DIR=%%~dp0.>> "%LUAJIT_CMD%"
  echo call %%TORCH_INSTALL_DIR%%\torch-activate.cmd>> "%LUAJIT_CMD%"
  echo %%TORCH_INSTALL_DIR%%\bin\luajit.exe %%*>> "%LUAJIT_CMD%"
  echo endlocal>> "%LUAJIT_CMD%"
)

if not "%LUA_CMD%" == "" (
  if exist "%LUA_CMD%" del %LUA_CMD%
  echo @echo off>> "%LUA_CMD%"
  echo setlocal>> "%LUA_CMD%"
  echo set TORCH_INSTALL_DIR=%%~dp0.>> "%LUA_CMD%"
  echo call %%TORCH_INSTALL_DIR%%\torch-activate.cmd>> "%LUA_CMD%"
  echo %%TORCH_INSTALL_DIR%%\bin\lua.exe %%*>> "%LUA_CMD%"
  echo endlocal>> "%LUA_CMD%"
)

if not "%LUAC_CMD%" == "" (
  if exist "%LUAC_CMD%" del %LUAC_CMD%
  echo @echo off>> "%LUAC_CMD%"
  echo setlocal>> "%LUAC_CMD%"
  echo set TORCH_INSTALL_DIR=%%~dp0.>> "%LUAC_CMD%"
  echo call %%TORCH_INSTALL_DIR%%\torch-activate.cmd>> "%LUAC_CMD%"
  echo %%TORCH_INSTALL_DIR%%\bin\luac.exe %%*>> "%LUAC_CMD%"
  echo endlocal>> "%LUAC_CMD%"
)

if exist %LUAROCKS_CMD% del %LUAROCKS_CMD%
echo @echo off>> %LUAROCKS_CMD%
echo setlocal>> %LUAROCKS_CMD%
echo set TORCH_INSTALL_DIR=%%~dp0.>> %LUAROCKS_CMD%
echo call %%TORCH_INSTALL_DIR%%\torch-activate.cmd>> %LUAROCKS_CMD%
echo call %%TORCH_INSTALL_DIR%%\luarocks\luarocks.bat %%*>> %LUAROCKS_CMD%
echo endlocal>> %LUAROCKS_CMD%

set CMAKE_CMD=%TORCH_INSTALL_DIR%\cmake.cmd
if exist %CMAKE_CMD% del %CMAKE_CMD%
echo @echo off>> %CMAKE_CMD%
echo if "%%1" == ".." if not "%%2" == "-G" goto :G_NMake>> %CMAKE_CMD%
echo cmake.exe %%*>> %CMAKE_CMD%
echo goto :EOF>> %CMAKE_CMD%
echo :G_NMake>> %CMAKE_CMD%
echo shift>> %CMAKE_CMD%
echo cmake.exe .. -G "NMake Makefiles" %%*>> %CMAKE_CMD%

set TORCH_SETUP_FAIL=0
cd %TORCH_DISTRO%
echo %ECHO_PREFIX% Setup succeed!
goto :END

:FAIL
cd %TORCH_DISTRO%
echo %ECHO_PREFIX% Setup fail!

:END



================================================
FILE: install-openblas.sh
================================================
#!/usr/bin/env bash
#OpenBLAS installation from source code
git clone https://github.com/xianyi/OpenBLAS.git ./OpenBLAS
cd OpenBLAS
make USE_OPENMP=1
mkdir ~/OpenBLAS
make install PREFIX=~/OpenBLAS
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:~/OpenBLAS/lib
SET_LLP="export LD_LIBRARY_PATH=\$LD_LIBRARY_PATH:~/OpenBLAS/lib"
echo "$SET_LLP" >> ~/.bashrc
cd ../



================================================
FILE: install.bat
================================================
@setlocal enableextensions
@echo off

::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
:::: This script instals Torch7 on windows with msvc        ::::
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::


if not "%TORCH_SETUP_FAIL%" == "0" call install-deps.bat
if not "%TORCH_SETUP_FAIL%" == "0" goto :FAIL

echo %ECHO_PREFIX% Updating submodules
git submodule update --init --recursive

set PATCH_DIR=%TORCH_DISTRO%\win-files\patch

echo %ECHO_PREFIX% Installing common lua packages
cd %TORCH_DISTRO%\extra\luafilesystem && call %LUAROCKS_CMD% make rockspecs\luafilesystem-1.6.3-1.rockspec || goto :FAIL
cd %TORCH_DISTRO%\extra\penlight && call %LUAROCKS_CMD% make penlight-scm-1.rockspec || goto :FAIL
cd %TORCH_DISTRO%\extra\lua-cjson && git apply %PATCH_DIR%\lua-cjson.patch --whitespace=fix & ( call %LUAROCKS_CMD% make lua-cjson-2.1devel-1.rockspec || goto :FAIL ) & git apply %PATCH_DIR%\lua-cjson.patch --reverse --whitespace=fix

echo %ECHO_PREFIX% Installing core Torch7 packages
cd %TORCH_DISTRO%\extra\luaffifb && git apply %PATCH_DIR%\luaffifb.patch --whitespace=fix & ( call %LUAROCKS_CMD% make luaffi-scm-1.rockspec || goto :FAIL ) & git apply %PATCH_DIR%\luaffifb.patch --reverse --whitespace=fix
cd %TORCH_DISTRO%\pkg\sundown && call %LUAROCKS_CMD% make rocks\sundown-scm-1.rockspec || goto :FAIL
cd %TORCH_DISTRO%\pkg\cwrap && call %LUAROCKS_CMD% make rocks\cwrap-scm-1.rockspec || goto :FAIL
cd %TORCH_DISTRO%\pkg\paths && call %LUAROCKS_CMD% make rocks\paths-scm-1.rockspec || goto :FAIL
if "%TORCH_SETUP_HAS_MKL%" == "1" (
  cd %TORCH_DISTRO%\pkg\torch && git apply %PATCH_DIR%\torch.patch --whitespace=fix & ( call %LUAROCKS_CMD% make rocks\torch-scm-1.rockspec INTEL_MKL_DIR="%INTEL_MKL_DIR%" INTEL_COMPILER_DIR="%INTEL_COMPILER_DIR%" || goto :FAIL ) & git apply %PATCH_DIR%\torch.patch --reverse --whitespace=fix
) else (
  cd %TORCH_DISTRO%\pkg\torch && git apply %PATCH_DIR%\torch.patch --whitespace=fix & ( call %LUAROCKS_CMD% make rocks\torch-scm-1.rockspec BLAS_LIBRARIES="%BLAS_LIBRARIES%" LAPACK_LIBRARIES="%LAPACK_LIBRARIES%" LAPACK_FOUND=TRUE || goto :FAIL ) & git apply %PATCH_DIR%\torch.patch --reverse --whitespace=fix
)
cd %TORCH_DISTRO%\pkg\dok && call %LUAROCKS_CMD% make rocks\dok-scm-1.rockspec || goto :FAIL
cd %TORCH_DISTRO%\exe\trepl && call %LUAROCKS_CMD% make trepl-scm-1.rockspec || goto :FAIL
cd %TORCH_DISTRO%\pkg\sys && call %LUAROCKS_CMD% make sys-1.1-0.rockspec || goto :FAIL
cd %TORCH_DISTRO%\pkg\xlua && call %LUAROCKS_CMD% make xlua-1.0-0.rockspec || goto :FAIL
cd %TORCH_DISTRO%\extra\nn && call %LUAROCKS_CMD% make rocks\nn-scm-1.rockspec || goto :FAIL
cd %TORCH_DISTRO%\extra\graph && call %LUAROCKS_CMD% make rocks\graph-scm-1.rockspec || goto :FAIL
cd %TORCH_DISTRO%\extra\nngraph && call %LUAROCKS_CMD% make nngraph-scm-1.rockspec || goto :FAIL
cd %TORCH_DISTRO%\pkg\image && call %LUAROCKS_CMD% make image-1.1.alpha-0.rockspec || goto :FAIL
cd %TORCH_DISTRO%\pkg\optim && call %LUAROCKS_CMD% make optim-1.0.5-0.rockspec || goto :FAIL

if not "%TORCH_SETUP_HAS_CUDA%" == "" if not "%TORCH_VS_TARGET%" == "x86" (
  echo %ECHO_PREFIX% Found CUDA on your machine. Installing CUDA packages
  cd %TORCH_DISTRO%\extra\cutorch && call %LUAROCKS_CMD% make rocks\cutorch-scm-1.rockspec || goto :FAIL
  cd %TORCH_DISTRO%\extra\cunn && call %LUAROCKS_CMD% make rocks\cunn-scm-1.rockspec || goto :FAIL
)

echo %ECHO_PREFIX% Installing optional Torch7 packages
cd %TORCH_DISTRO%\pkg\gnuplot && call %LUAROCKS_CMD% make rocks\gnuplot-scm-1.rockspec
cd %TORCH_DISTRO%\exe\env && call %LUAROCKS_CMD% make env-scm-1.rockspec
cd %TORCH_DISTRO%\extra\nnx && call %LUAROCKS_CMD% make nnx-0.1-1.rockspec
cd %TORCH_DISTRO%\exe\qtlua && call %LUAROCKS_CMD% make rocks\qtlua-scm-1.rockspec
cd %TORCH_DISTRO%\pkg\qttorch && call %LUAROCKS_CMD% make rocks\qttorch-scm-1.rockspec
cd %TORCH_DISTRO%\extra\threads && call %LUAROCKS_CMD% make rocks\threads-scm-1.rockspec WIN_DLFCN_INCDIR=%WIN_DLFCN_INCDIR% WIN_DLFCN_LIBDIR=%WIN_DLFCN_LIBDIR%
cd %TORCH_DISTRO%\extra\argcheck && call %LUAROCKS_CMD% make rocks\argcheck-scm-1.rockspec

if not "%TORCH_SETUP_HAS_CUDA%" == "" if not "%TORCH_VS_TARGET%" == "x86" (
  echo %ECHO_PREFIX% Found CUDA on your machine. Installing optional CUDA packages
  cd %TORCH_DISTRO%\extra\cudnn && call %LUAROCKS_CMD% make cudnn-scm-1.rockspec
)

echo %ECHO_PREFIX% Installation succeed!
goto :END

:FAIL
echo %ECHO_PREFIX% Installation error!

:END
@endlocal



================================================
FILE: install.sh
================================================
#!/usr/bin/env bash

export CC=
export CXX=

#compiling option

intel=$1    # compiler icc/gcc
avx512=$2   # AVX512F on/off 
#omp=$3      # intel intel/gnu
use_mklml=$3      # intel intel/gnu
skip=$4     # skip openblas, value = noskip/skip, default is noskip
if [[ $skip == 'skip' ]]; then
   echo "skip openblas check"
else
   #identify user level, and check openblas.For root user, check /opt/OpenBLAS, for none root user, check ~/OpenBLAS
   if [ "$(ls -A /opt/OpenBLAS)" ]; then
      echo "OpenBLAS is installed in /opt"
   else
      if [[ $EUID -ne 0 ]]; then
         #none root user
         RTN=$(echo $LD_LIBRARY_PATH |grep -i openblas)
         if [ -z "$RTN" ]; then
            echo "OpenBLAS is not installed, or environment is not set"
            bash install-openblas.sh
         else
            echo "OpenBLAS installed."
   
         fi
      else
         #root user
         echo "ERROR: please instal OpenBLAS: bash install-deps , or apt-get install -y libopenblas-dev "
      fi
   fi
fi


FORCE_AVX512_v=OFF
if [[ $avx512 == 'avx512' ]]; then
  FORCE_AVX512_v=ON
else
  FORCE_AVX512_v=OFF
fi

USE_MKLML_v=OFF
if [[ $use_mklml == 'mklml' ]]; then
  USE_MKLML_v=ON
else
  USE_MKLML_v=OFF
fi

WITH_IOMP_v=ON
:<<hallo
if [[ $omp == 'iomp' ]]; then
  WITH_IOMP_v=ON
elif [[ $omp == 'gomp' ]]; then 
  WITH_IOMP_v=OFF
fi
hallo

SKIP_R=0
BATCH_INSTALL=0


THIS_DIR=$(cd $(dirname $0); pwd)
if [[ "$THIS_DIR" == *" "* ]]; then
    echo "$THIS_DIR: Torch cannot install to a path containing whitespace.
Please try a different path, one without any spaces."
    exit 1
fi
PREFIX=${PREFIX:-"${THIS_DIR}/install"}
TORCH_LUA_VERSION=${TORCH_LUA_VERSION:-"LUAJIT21"} # by default install LUAJIT21

while getopts 'bsh' x; do
    case "$x" in
        h)
            echo "usage: $0
This script will install Torch and related, useful packages into $PREFIX.

    -b      Run without requesting any user input (will automatically add PATH to shell profile)
    -s      Skip adding the PATH to shell profile
"
            exit 2
            ;;
        b)
            BATCH_INSTALL=1
            ;;
        s)
            SKIP_RC=1
            ;;
    esac
done


# Scrub an anaconda/conda install, if exists, from the PATH.
# It has a malformed MKL library (as of 1/17/2015)
OLDPATH=$PATH
if [[ $(echo $PATH | grep conda) ]]; then
    export PATH=$(echo $PATH | tr ':' '\n' | grep -v "conda[2-9]\?" | uniq | tr '\n' ':')
fi

echo "Prefix set to $PREFIX"

if [[ `uname` == 'Linux' ]]; then
    ICC_ON=0
    if [[ $intel == 'icc' ]]; then
        if [[ `which icc` == '' ]]; then
            CC=gcc
        else
            CC=icc
        fi    
    
        if [[ `which icpc` == '' ]]; then
            CXX=g++
        else
            CXX=icpc
        fi   
        if [[ `echo $CC | grep icc` != '' ]]; then
            ICC_ON=1
        fi
    else
        ICC_ON=0
    fi
    
    if [[ $ICC_ON == 1 ]]; then
       echo 'using Intel compiler to compile torch'
    else
       echo 'using GNU compilter to compile torch'
    fi

    #`./prepare_mklml.sh $ICC_ON`
    RETURN_STRING=`./prepare_mklml.sh $ICC_ON $USE_MKLML_v`
    export MKLML_ROOT=`echo $RETURN_STRING | awk '{print $1}'`
    if [ $USE_MKLML_v == ON ]; then
      echo "Path of MKLML_ROOT is:"
    else
      echo "Path of MKL_ROOT is:"
    fi
    echo $MKLML_ROOT
    MKLML_LIBRARY_PATH=$MKLROOT/lib
    MKLML_INCLUDE_PATH=$MKLROOT/include
    
    export CMAKE_LIBRARY_PATH=/opt/OpenBLAS/include:/opt/OpenBLAS/lib:$MKLML_LIBRARY_PATH:$CMAKE_LIBRARY_PATH
    export CMAKE_INCLUDE_PATH=/opt/OpenBLAS/include:$MKLML_INCLUDE_PATH:$CMAKE_INCLUDE_PATH
    export CMAKE_LIBRARY_PATH=$PREFIX/include:/opt/OpenBLAS/include:$PREFIX/lib:/opt/OpenBLAS/lib:$CMAKE_LIBRARY_PATH
    export CMAKE_LIBRARY_PATH=~/OpenBLAS/include:~/OpenBLAS/lib:$CMAKE_LIBRARY_PATH
    export LD_LIBRARY_PATH=~/OpenBLAS/include:~/OpenBLAS/lib:$LD_LIBRARY_PATH

fi

export CMAKE_PREFIX_PATH=$PREFIX
if [[ $intel == 'gcc' ]]; then
  export CC=`which gcc`
  export CXX=`which g++`
fi

git submodule update --init --recursive

# If we're on OS X, use clang
if [[ `uname` == "Darwin" ]]; then
    # make sure that we build with Clang. CUDA's compiler nvcc
    # does not play nice with any recent GCC version.
    export CC=clang
    export CXX=clang++
fi
# If we're on Arch linux, use gcc v5
if [[ `uname -a` == *"ARCH"* ]]; then
    path_to_gcc5=$(which gcc-5)
    if [ -x "$path_to_gcc5" ]; then
      export CC="$path_to_gcc5"
    else
      echo "Warning: GCC v5 not found. CUDA v8 is incompatible with GCC v6, if installation fails, consider running \$ pacman -S gcc5"
    fi
fi

echo "Installing Lua version: ${TORCH_LUA_VERSION}"
mkdir -p install
mkdir -p build
cd build
cmake .. -DCMAKE_INSTALL_PREFIX="${PREFIX}" -DCMAKE_BUILD_TYPE=Release -DWITH_${TORCH_LUA_VERSION}=ON 2>&1 >>$PREFIX/install.log || exit 1
(make 2>&1 >>$PREFIX/install.log  || exit 1) && (make install 2>&1 >>$PREFIX/install.log || exit 1)
cd ..

# Check for a CUDA install (using nvcc instead of nvidia-smi for cross-platform compatibility)
path_to_nvcc=$(which nvcc)
if [ $? == 1 ]; then { # look for it in /usr/local
  if [[ -f /usr/local/cuda/bin/nvcc ]]; then {
    path_to_nvcc=/usr/local/cuda/bin/nvcc
  } 
  elif [[ -f /opt/cuda/bin/nvcc ]]; then { # default path for arch
    path_to_nvcc=/opt/cuda/bin/nvcc
  } fi
} fi

# check if we are on mac and fix RPATH for local install
path_to_install_name_tool=$(which install_name_tool 2>/dev/null)
if [ -x "$path_to_install_name_tool" ]
then
   if [ ${TORCH_LUA_VERSION} == "LUAJIT21" ] || [ ${TORCH_LUA_VERSION} == "LUAJIT20" ] ; then
       install_name_tool -id ${PREFIX}/lib/libluajit.dylib ${PREFIX}/lib/libluajit.dylib
   else
       install_name_tool -id ${PREFIX}/lib/liblua.dylib ${PREFIX}/lib/liblua.dylib
   fi
fi

if [ -x "$path_to_nvcc" ] || [ -x "$path_to_nvidiasmi" ]
then
    echo "Found CUDA on your machine. Installing CMake 3.6 modules to get up-to-date FindCUDA"
    cd ${THIS_DIR}/cmake/3.6 && \
(cmake -E make_directory build && cd build && cmake .. -DCMAKE_INSTALL_PREFIX="${PREFIX}" \
        && make install) && echo "FindCuda bits of CMake 3.6 installed" || exit 1
fi

setup_lua_env_cmd=$($PREFIX/bin/luarocks path)
eval "$setup_lua_env_cmd"

echo "$CC"
echo "Installing common Lua packages"
cd ${THIS_DIR}/extra/luafilesystem && $PREFIX/bin/luarocks make rockspecs/luafilesystem-1.6.3-1.rockspec || exit 1
cd ${THIS_DIR}/extra/penlight && $PREFIX/bin/luarocks make penlight-scm-1.rockspec || exit 1
cd ${THIS_DIR}/extra/lua-cjson && $PREFIX/bin/luarocks make lua-cjson-2.1devel-1.rockspec || exit 1

echo "Installing core Torch packages"
cd ${THIS_DIR}/extra/luaffifb && $PREFIX/bin/luarocks make luaffi-scm-1.rockspec       || exit 1
cd ${THIS_DIR}/pkg/sundown   && $PREFIX/bin/luarocks make rocks/sundown-scm-1.rockspec || exit 1
cd ${THIS_DIR}/pkg/cwrap        && $PREFIX/bin/luarocks make rocks/cwrap-scm-1.rockspec   || exit 1
cd ${THIS_DIR}/pkg/paths        && $PREFIX/bin/luarocks make rocks/paths-scm-1.rockspec   || exit 1
cd ${THIS_DIR}/pkg/torch        && $PREFIX/bin/luarocks make WITH_IOMP=$WITH_IOMP_v FORCE_AVX512=$FORCE_AVX512_v USE_MKLML=$USE_MKLML_v rocks/torch-scm-1.rockspec   || exit 1
cd ${THIS_DIR}/pkg/dok          && $PREFIX/bin/luarocks make rocks/dok-scm-1.rockspec     || exit 1
cd ${THIS_DIR}/exe/trepl        && $PREFIX/bin/luarocks make trepl-scm-1.rockspec         || exit 1
cd ${THIS_DIR}/pkg/sys          && $PREFIX/bin/luarocks make sys-1.1-0.rockspec           || exit 1
cd ${THIS_DIR}/pkg/xlua         && $PREFIX/bin/luarocks make xlua-1.0-0.rockspec          || exit 1
cd ${THIS_DIR}/extra/nn         && $PREFIX/bin/luarocks make FORCE_AVX512=$FORCE_AVX512_v rocks/nn-scm-1.rockspec      || exit 1
cd ${THIS_DIR}/extra/mkltorch   && $PREFIX/bin/luarocks make FORCE_AVX512=$FORCE_AVX512_v mkltorch-scm-1.rockspec      || exit 1
cd ${THIS_DIR}/extra/mklnn      && $PREFIX/bin/luarocks make FORCE_AVX512=$FORCE_AVX512_v mklnn-scm-1.rockspec      || exit 1
cd ${THIS_DIR}/extra/graph      && $PREFIX/bin/luarocks make rocks/graph-scm-1.rockspec   || exit 1
cd ${THIS_DIR}/extra/nngraph    && $PREFIX/bin/luarocks make nngraph-scm-1.rockspec       || exit 1
cd ${THIS_DIR}/pkg/image        && $PREFIX/bin/luarocks make image-1.1.alpha-0.rockspec   || exit 1
cd ${THIS_DIR}/pkg/optim        && $PREFIX/bin/luarocks make optim-1.0.5-0.rockspec       || exit 1

if [ -x "$path_to_nvcc" ]
then
    echo "Found CUDA on your machine. Installing CUDA packages"
    cd ${THIS_DIR}/extra/cutorch && $PREFIX/bin/luarocks make rocks/cutorch-scm-1.rockspec || exit 1
    cd ${THIS_DIR}/extra/cunn    && $PREFIX/bin/luarocks make rocks/cunn-scm-1.rockspec    || exit 1
fi

# Optional packages
echo "Installing optional Torch packages"
cd ${THIS_DIR}/pkg/gnuplot          && $PREFIX/bin/luarocks make rocks/gnuplot-scm-1.rockspec
cd ${THIS_DIR}/exe/env              && $PREFIX/bin/luarocks make env-scm-1.rockspec
cd ${THIS_DIR}/extra/nnx            && $PREFIX/bin/luarocks make nnx-0.1-1.rockspec
cd ${THIS_DIR}/exe/qtlua            && $PREFIX/bin/luarocks make rocks/qtlua-scm-1.rockspec
cd ${THIS_DIR}/pkg/qttorch          && $PREFIX/bin/luarocks make rocks/qttorch-scm-1.rockspec
cd ${THIS_DIR}/extra/threads        && $PREFIX/bin/luarocks make rocks/threads-scm-1.rockspec
cd ${THIS_DIR}/extra/argcheck       && $PREFIX/bin/luarocks make rocks/argcheck-scm-1.rockspec

# Optional CUDA packages
if [ -x "$path_to_nvcc" ]
then
    echo "Found CUDA on your machine. Installing optional CUDA packages"
    cd ${THIS_DIR}/extra/cudnn   && $PREFIX/bin/luarocks make cudnn-scm-1.rockspec
fi

export PATH=$OLDPATH # Restore anaconda distribution if we took it out.

# Add C libs to LUA_CPATH
if [[ `uname` == "Darwin" ]]; then
    CLIB_LUA_CPATH=$PREFIX/lib/?.dylib
else
    CLIB_LUA_CPATH=$PREFIX/lib/?.so
fi

cat <<EOF >$PREFIX/bin/torch-activate
$setup_lua_env_cmd
export PATH=$PREFIX/bin:\$PATH
export LD_LIBRARY_PATH=$PREFIX/lib:\$LD_LIBRARY_PATH
export DYLD_LIBRARY_PATH=$PREFIX/lib:\$DYLD_LIBRARY_PATH
export LUA_CPATH='$CLIB_LUA_CPATH;'\$LUA_CPATH
EOF
chmod +x $PREFIX/bin/torch-activate

if [[ $SKIP_RC == 1 ]]; then
  exit 0
fi

RC_FILE=0
DEFAULT=yes
if [[ $(echo $SHELL | grep bash) ]]; then
    RC_FILE=$HOME/.bashrc
elif [[ $(echo $SHELL | grep zsh) ]]; then
    RC_FILE=$HOME/.zshrc
else
    echo "

Non-standard shell $SHELL detected. You might want to
add the following lines to your shell profile:

. $PREFIX/bin/torch-activate
"
fi

if [[ $skip == 'skip' ]]; then
   BATCH_INSTALL=1
   RC_FILE=
fi

WRITE_PATH_TO_PROFILE=0
if [[ $BATCH_INSTALL == 0 ]]; then
    if [ -f "$RC_FILE" ]; then
        echo "

Do you want to automatically prepend the Torch install location
to PATH and LD_LIBRARY_PATH in your $RC_FILE? (yes/no)
[$DEFAULT] >>> "
        read input
        if [[ $input == "" ]]; then
            input=$DEFAULT
        fi

        is_yes() {
            yesses={y,Y,yes,Yes,YES}
            if [[ $yesses =~ $1 ]]; then
                echo 1
            fi
        }

        if [[ $(is_yes $input) ]]; then
            WRITE_PATH_TO_PROFILE=1
        fi
    fi
else
    if [[ "$RC_FILE" ]]; then
        WRITE_PATH_TO_PROFILE=1
    fi
fi

WRITE_PATH_TO_PROFILE=0
if [[ $WRITE_PATH_TO_PROFILE == 1 ]]; then
    echo "

. $PREFIX/bin/torch-activate" >> "$RC_FILE"
    echo "

. $PREFIX/bin/torch-activate" >> "$HOME"/.profile

else
    echo "

Not updating your shell profile.
You might want to
add the following lines to your shell profile:

. $PREFIX/bin/torch-activate
"
fi



================================================
FILE: LICENSE.md
================================================
Copyright (c) 2016, Soumith Chintala, Ronan Collobert, Koray Kavukcuoglu, Clement Farabet
All rights reserved.

Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:

* Redistributions of source code must retain the above copyright notice, this
  list of conditions and the following disclaimer.

* Redistributions in binary form must reproduce the above copyright notice,
  this list of conditions and the following disclaimer in the documentation
  and/or other materials provided with the distribution.

* Neither the name of distro nor the names of its
  contributors may be used to endorse or promote products derived from
  this software without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.



================================================
FILE: prepare_mklml.sh
================================================
#!/bin/sh
# set -ex
FindLibrary() 
{
  case "$1" in
    intel|1)
      LOCALMKLML=`find $DST -name libmklml_intel.so`   # name of MKL SDL lib
      ;;
    *)
      LOCALMKLML=`find $DST -name libmklml_gnu.so`   # name of MKL SDL lib
      ;;
  esac

}

GetVersionName()
{
VERSION_LINE=0
if [ $1 ]; then
  VERSION_LINE=`grep __INTEL_MKL_BUILD_DATE $1/include/mkl_version.h 2>/dev/null | sed -e 's/.* //'`
fi
if [ -z $VERSION_LINE ]; then
  VERSION_LINE=0
fi
echo $VERSION_LINE  # Return Version Line
}

# MKLML
DST=`dirname $0`
OMP=0 
VERSION_MATCH=20170908
ARCHIVE_BASENAME=mklml_lnx_2018.0.1.20171007.tgz
MKLML_CONTENT_DIR=`echo $ARCHIVE_BASENAME | rev | cut -d "." -f 2- | rev`
GITHUB_RELEASE_TAG=v0.11
MKLMLURL="https://github.com/01org/mkl-dnn/releases/download/$GITHUB_RELEASE_TAG/$ARCHIVE_BASENAME"
# there are diffrent MKL lib to be used for GCC and for ICC
reg='^[0-9]+$'
#echo "----->"  $DST
VERSION_LINE=`GetVersionName $MKLMLROOT`
if [[ $2 == ON ]]; then
  # Check if MKLMLROOT is set if positive then set one will be used..
  if [ -z $MKLMLROOT ] || [ $VERSION_LINE -lt $VERSION_MATCH ]; then
  	# ..if MKLROOT is not set then check if we have MKL downloaded in proper version
      VERSION_LINE=`GetVersionName $DST/$MKLML_CONTENT_DIR`
      if [ $VERSION_LINE -lt $VERSION_MATCH ] ; then
        #...If it is not then downloaded and unpacked
        wget --no-check-certificate -P $DST $MKLMLURL -O $DST/$ARCHIVE_BASENAME
        tar -xzf $DST/$ARCHIVE_BASENAME -C $DST
      fi
  
    FindLibrary $1
    #echo "----- $LOCALMKLML"
    MKLMLROOT=$PWD/`echo $LOCALMKLML | sed -e 's/lib.*$//'`
  fi
  # return value to calling script (Makefile,cmake)
  echo $MKLMLROOT $LIBRARIES $OMP

else 
  
  # Check what MKL lib we have in MKLROOT

  if [ -z `find $MKLROOT -name libmkl_rt.so -print -quit` ]; then
    LIBRARIES=`basename $LOCALMKL | sed -e 's/^.*lib//' | sed -e 's/\.so.*$//'`
    OMP=1
  else
    LIBRARIES="mkl_rt"
  fi 
  echo $MKLROOT $LIBRARIES $OMP

fi




================================================
FILE: test.bat
================================================
@setlocal enableextensions
@echo off

::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
:::: This script tests if Torch7 is installed properly      ::::
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::


if "%TORCH_INSTALL_DIR%" == "" goto :HELP

for /f "delims=" %%i in ('where luajit.cmd') do (
  set LUA=%%i
  goto :AFTER_LUAJIT
)

:AFTER_LUAJIT
if exist "%LUA%" (
  set TORCH_USE_LUAJIT=1
) else (
  for /f "delims=" %%i in ('where lua.cmd') do (
    set LUA=%%i
    goto :AFTER_LUA
  )
)

:AFTER_LUA
if not exist "%LUA%" (
  echo Neither luajit nor lua found in Torch7 environment
  goto :FAIL
)

set LUA_SAFE_PATH=%LUA:\=\\%
echo Using Lua at: %LUA%

:: smoke tests
call %LUA% -lpaths     -e "print('paths loaded succesfully')" || goto :FAIL
call %LUA% -ltorch     -e "print('torch loaded succesfully')" || goto :FAIL
call %LUA% -lenv       -e "print('env loaded succesfully')" || goto :FAIL
call %LUA% -ltrepl     -e "print('trepl loaded succesfully')" || goto :FAIL
call %LUA% -ldok       -e "print('dok loaded succesfully')" || goto :FAIL
call %LUA% -limage     -e "print('image loaded succesfully')" || goto :FAIL
call %LUA% -lcwrap     -e "print('cwrap loaded succesfully')" || goto :FAIL
call %LUA% -lgnuplot   -e "print('gnuplot loaded succesfully')" || goto :FAIL
call %LUA% -loptim     -e "print('optim loaded succesfully')" || goto :FAIL
call %LUA% -lsys       -e "print('sys loaded succesfully')" || goto :FAIL
for /f "delims=" %%i in ('where basename') do set BASENAME=%%i
if "%BASENAME%" == "" (
  call %LUA% -lxlua    -e "print('xlua loaded succesfully')" || goto :FAIL
) else (
  call %LUA% -lxlua    -e "print('x$(basename %LUA_SAFE_PATH%) loaded succesfully')" || goto :FAIL
)
call %LUA% -largcheck  -e "print('argcheck loaded succesfully')" || goto :FAIL
call %LUA% -lgraph     -e "print('graph loaded succesfully')" || goto :FAIL
call %LUA% -lnn        -e "print('nn loaded succesfully')" || goto :FAIL
call %LUA% -lnngraph   -e "print('nngraph loaded succesfully')" || goto :FAIL
call %LUA% -lnnx       -e "print('nnx loaded succesfully')" || goto :FAIL
call %LUA% -lthreads   -e "print('threads loaded succesfully')" || goto :FAIL

call th -ltorch -e "torch.test()"
call th -lnn    -e "nn.test()"

if "%TORCH_USE_LUAJIT%" == "1" (
  call %LUA% -lsundown -e "print('sundown loaded succesfully')"
)

call %LUA% -lcutorch -e ""
if "%ERRORLEVEL%" == "0" (
  call %LUA% -lcutorch -e "print('cutorch loaded succesfully')"
  call %LUA% -lcunn    -e "print('cunn loaded succesfully')"
  if "%TORCH_USE_LUAJIT%" == "1" %LUA% -lcudnn -e "print('cudnn loaded succesfully')"
  call th -lcutorch    -e "cutorch.test()"
  call th -lcunn       -e "nn.testcuda()"
) else (
  echo "CUDA not found"
)

echo Test succeed
goto :END

:HELP
echo Please run torch-activate.cmd before run test.bat so that test knows where to find Torch7.
goto :END

:FAIL
echo Test fail

:END
@endlocal



================================================
FILE: test.sh
================================================
#!/bin/bash
set -e

LUA=$(which luajit lua | head -n 1)

if [ ! -x "$LUA" ]
then
    echo "Neither luajit nor lua found in path"
    exit 1
fi

echo "Using Lua at:"
echo "$LUA"

#smoke tests
$LUA -lpaths     -e "print('paths loaded succesfully')"
$LUA -ltorch     -e "print('torch loaded succesfully')"
$LUA -lenv       -e "print('env loaded succesfully')"
$LUA -ltrepl     -e "print('trepl loaded succesfully')"
$LUA -ldok       -e "print('dok loaded succesfully')"
$LUA -limage     -e "print('image loaded succesfully')"
$LUA -lcwrap     -e "print('cwrap loaded succesfully')"
$LUA -lgnuplot   -e "print('gnuplot loaded succesfully')"
$LUA -loptim     -e "print('optim loaded succesfully')"
$LUA -lsys       -e "print('sys loaded succesfully')"
$LUA -lxlua      -e "print('x$(basename $LUA) loaded succesfully')"
$LUA -largcheck  -e "print('argcheck loaded succesfully')"
$LUA -lgraph     -e "print('graph loaded succesfully')"
$LUA -lnn        -e "print('nn loaded succesfully')"
$LUA -lmkltorch  -e "print('mkltorch loaded succesfully')"
$LUA -lmklnn     -e "print('mklnn loaded succesfully')"
$LUA -lnngraph   -e "print('nngraph loaded succesfully')"
$LUA -lnnx       -e "print('nnx loaded succesfully')"
$LUA -lthreads   -e "print('threads loaded succesfully')"

th -ltorch -e "torch.test()"
th -lnn    -e "nn.test()"
th -lmklnn -e "mklnn.test()"

if [ $(basename $LUA) = "luajit" ]
then
    $LUA -lsundown         -e "print('sundown loaded succesfully')"
fi

if `$LUA -lcutorch -e ""`
then
    $LUA -lcutorch -e "print('cutorch loaded succesfully')"
    $LUA -lcunn -e "print('cunn loaded succesfully')"
    if [ $(basename $LUA) = "luajit" ];
    then
        $LUA -lcudnn -e "print('cudnn loaded succesfully')"
    fi
    th -lcutorch -e "cutorch.test()"
    th -lcunn -e "nn.testcuda()"
else
    echo "CUDA not found"
fi



================================================
FILE: travis_cuda_install.sh
================================================
#!/usr/bin/env sh

if [[ `uname` == 'Linux' ]]; then
  CUDA_VERSION=6-5
  CUDA_URL=http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1204/x86_64/cuda-repo-ubuntu1204_6.5-14_amd64.deb
  CUDA_FILE=/tmp/cuda_install.deb

  curl $CUDA_URL -o $CUDA_FILE
  dpkg -i $CUDA_FILE
  rm -f $CUDA_FILE
  apt-get -y update
  apt-get -y install \
    cuda-core-${CUDA_VERSION} \
    cuda-cublas-${CUDA_VERSION} \
    cuda-cublas-dev-${CUDA_VERSION} \
    cuda-cudart-${CUDA_VERSION} \
    cuda-cudart-dev-${CUDA_VERSION} \
    cuda-curand-${CUDA_VERSION} \
    cuda-curand-dev-${CUDA_VERSION} \
    cuda-cusparse-${CUDA_VERSION} \
    cuda-cusparse-dev-${CUDA_VERSION}
  ln -s /usr/local/cuda-6.5 /usr/local/cuda
fi



================================================
FILE: uninstall.bat
================================================
@setlocal enableextensions
@echo off

::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
:::: This script [1] cleans temporary compilation file      ::::
::::             [2] deletes Torch7 installation directory  ::::
::::             [3] deletes Torch7 conda environment       ::::
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::


if "%TORCH_INSTALL_DIR%" == "" goto :HELP

set ECHO_PREFIX=+++++++

echo %ECHO_PREFIX% cleaning temporary compilation files
call clean.bat

echo %ECHO_PREFIX% deleting Torch7 installation directory
rmdir /s /q %TORCH_INSTALL_DIR%

echo %ECHO_PREFIX% deleting Torch7 conda environment (even if env is removed, packages will stil be kept in conda/pkgs)
conda env remove -n %TORCH_CONDA_ENV% --yes

echo %ECHO_PREFIX% Torch7 has been uninstalled
goto :END

:HELP
echo Please run torch-activate.cmd before run uninstall.bat so that uninstall knows where to find Torch7.

:END
@endlocal



================================================
FILE: update.sh
================================================
#!/usr/bin/env bash

git fetch
git reset --hard origin/master
# Submodule update is done inside install.sh
./install.sh -s



================================================
FILE: .travis.yml
================================================
language: c
compiler:
  - gcc
  - clang
cache:
  directories:
  - $HOME/OpenBlasInstall
sudo: required
env:
  - TORCH_LUA_VERSION=LUAJIT21
  - TORCH_LUA_VERSION=LUA51
  - TORCH_LUA_VERSION=LUA52
addons:
  apt:
    packages:
    - cmake
    - gfortran
    - gcc-multilib
    - gfortran-multilib
    - liblapack-dev
    - build-essential
    - gcc 
    - g++ 
    - curl
    - cmake 
    - libreadline-dev 
    - git-core 
    - libqt4-core 
    - libqt4-gui
    - libqt4-dev 
    - libjpeg-dev 
    - libpng-dev 
    - ncurses-dev
    - imagemagick 
    - libzmq3-dev 
    - gfortran 
    - unzip 
    - gnuplot
    - gnuplot-x11 
    - libfftw3-dev 
    - sox 
    - libsox-dev
    - libsox-fmt-all
    - libgraphicsmagick1-dev
    - graphicsmagick
before_script: 
- export ROOT_TRAVIS_DIR=$(pwd)
- export INSTALL_PREFIX=~/torch/install
-  ls $HOME/OpenBlasInstall/lib || (cd /tmp/ && git clone https://github.com/xianyi/OpenBLAS.git -b master && cd OpenBLAS && (make NO_AFFINITY=1 -j$(getconf _NPROCESSORS_ONLN) 2>/dev/null >/dev/null) && make PREFIX=$HOME/OpenBlasInstall install)
- git clone https://github.com/torch/distro.git ~/torch --recursive
- sudo -E $ROOT_TRAVIS_DIR/travis_cuda_install.sh
script: 
- export PATH=/usr/local/cuda/bin/:$PATH
- cd ~/torch && ./install.sh -b
- source ~/torch/install/bin/torch-activate
- ./test.sh



================================================
FILE: admin/updateSubmodules.sh
================================================
git pull
git submodule update
# dont update luajit-rocks because of https://github.com/LuaJIT/LuaJIT/issues/325
git submodule foreach bash -c 'if [ $(basename $(pwd)) != 'luajit-rocks' ]; then git pull origin master; fi'
# git submodule foreach git pull origin master
git add extra pkg exe
git commit -m "updating packages"



================================================
FILE: cmake/FindTorch.cmake
================================================
# We are in the build tree, nothing to do
# We found ourselves :)



================================================
FILE: cmake/3.6/CMakeLists.txt
================================================
cmake_minimum_required(VERSION 2.8)

#
#  This CMake file installs some module files from CMake 3.6
#  into Torch cmake install directory.
#

INSTALL(DIRECTORY "${CMAKE_CURRENT_SOURCE_DIR}/Modules/"
        DESTINATION "share/cmake/torch")



================================================
FILE: cmake/3.6/Modules/FindCUDA.cmake
================================================
#.rst:
# FindCUDA
# --------
#
# Tools for building CUDA C files: libraries and build dependencies.
#
# This script locates the NVIDIA CUDA C tools.  It should work on linux,
# windows, and mac and should be reasonably up to date with CUDA C
# releases.
#
# This script makes use of the standard find_package arguments of
# <VERSION>, REQUIRED and QUIET.  CUDA_FOUND will report if an
# acceptable version of CUDA was found.
#
# The script will prompt the user to specify CUDA_TOOLKIT_ROOT_DIR if
# the prefix cannot be determined by the location of nvcc in the system
# path and REQUIRED is specified to find_package().  To use a different
# installed version of the toolkit set the environment variable
# CUDA_BIN_PATH before running cmake (e.g.
# CUDA_BIN_PATH=/usr/local/cuda1.0 instead of the default
# /usr/local/cuda) or set CUDA_TOOLKIT_ROOT_DIR after configuring.  If
# you change the value of CUDA_TOOLKIT_ROOT_DIR, various components that
# depend on the path will be relocated.
#
# It might be necessary to set CUDA_TOOLKIT_ROOT_DIR manually on certain
# platforms, or to use a cuda runtime not installed in the default
# location.  In newer versions of the toolkit the cuda library is
# included with the graphics driver- be sure that the driver version
# matches what is needed by the cuda runtime version.
#
# The following variables affect the behavior of the macros in the
# script (in alphebetical order).  Note that any of these flags can be
# changed multiple times in the same directory before calling
# CUDA_ADD_EXECUTABLE, CUDA_ADD_LIBRARY, CUDA_COMPILE, CUDA_COMPILE_PTX,
# CUDA_COMPILE_FATBIN, CUDA_COMPILE_CUBIN or CUDA_WRAP_SRCS::
#
#   CUDA_64_BIT_DEVICE_CODE (Default matches host bit size)
#   -- Set to ON to compile for 64 bit device code, OFF for 32 bit device code.
#      Note that making this different from the host code when generating object
#      or C files from CUDA code just won't work, because size_t gets defined by
#      nvcc in the generated source.  If you compile to PTX and then load the
#      file yourself, you can mix bit sizes between device and host.
#
#   CUDA_ATTACH_VS_BUILD_RULE_TO_CUDA_FILE (Default ON)
#   -- Set to ON if you want the custom build rule to be attached to the source
#      file in Visual Studio.  Turn OFF if you add the same cuda file to multiple
#      targets.
#
#      This allows the user to build the target from the CUDA file; however, bad
#      things can happen if the CUDA source file is added to multiple targets.
#      When performing parallel builds it is possible for the custom build
#      command to be run more than once and in parallel causing cryptic build
#      errors.  VS runs the rules for every source file in the target, and a
#      source can have only one rule no matter how many projects it is added to.
#      When the rule is run from multiple targets race conditions can occur on
#      the generated file.  Eventually everything will get built, but if the user
#      is unaware of this behavior, there may be confusion.  It would be nice if
#      this script could detect the reuse of source files across multiple targets
#      and turn the option off for the user, but no good solution could be found.
#
#   CUDA_BUILD_CUBIN (Default OFF)
#   -- Set to ON to enable and extra compilation pass with the -cubin option in
#      Device mode. The output is parsed and register, shared memory usage is
#      printed during build.
#
#   CUDA_BUILD_EMULATION (Default OFF for device mode)
#   -- Set to ON for Emulation mode. -D_DEVICEEMU is defined for CUDA C files
#      when CUDA_BUILD_EMULATION is TRUE.
#
#   CUDA_GENERATED_OUTPUT_DIR (Default CMAKE_CURRENT_BINARY_DIR)
#   -- Set to the path you wish to have the generated files placed.  If it is
#      blank output files will be placed in CMAKE_CURRENT_BINARY_DIR.
#      Intermediate files will always be placed in
#      CMAKE_CURRENT_BINARY_DIR/CMakeFiles.
#
#   CUDA_HOST_COMPILATION_CPP (Default ON)
#   -- Set to OFF for C compilation of host code.
#
#   CUDA_HOST_COMPILER (Default CMAKE_C_COMPILER, $(VCInstallDir)/bin for VS)
#   -- Set the host compiler to be used by nvcc.  Ignored if -ccbin or
#      --compiler-bindir is already present in the CUDA_NVCC_FLAGS or
#      CUDA_NVCC_FLAGS_<CONFIG> variables.  For Visual Studio targets
#      $(VCInstallDir)/bin is a special value that expands out to the path when
#      the command is run from within VS.
#
#   CUDA_NVCC_FLAGS
#   CUDA_NVCC_FLAGS_<CONFIG>
#   -- Additional NVCC command line arguments.  NOTE: multiple arguments must be
#      semi-colon delimited (e.g. --compiler-options;-Wall)
#
#   CUDA_PROPAGATE_HOST_FLAGS (Default ON)
#   -- Set to ON to propagate CMAKE_{C,CXX}_FLAGS and their configuration
#      dependent counterparts (e.g. CMAKE_C_FLAGS_DEBUG) automatically to the
#      host compiler through nvcc's -Xcompiler flag.  This helps make the
#      generated host code match the rest of the system better.  Sometimes
#      certain flags give nvcc problems, and this will help you turn the flag
#      propagation off.  This does not affect the flags supplied directly to nvcc
#      via CUDA_NVCC_FLAGS or through the OPTION flags specified through
#      CUDA_ADD_LIBRARY, CUDA_ADD_EXECUTABLE, or CUDA_WRAP_SRCS.  Flags used for
#      shared library compilation are not affected by this flag.
#
#   CUDA_SEPARABLE_COMPILATION (Default OFF)
#   -- If set this will enable separable compilation for all CUDA runtime object
#      files.  If used outside of CUDA_ADD_EXECUTABLE and CUDA_ADD_LIBRARY
#      (e.g. calling CUDA_WRAP_SRCS directly),
#      CUDA_COMPUTE_SEPARABLE_COMPILATION_OBJECT_FILE_NAME and
#      CUDA_LINK_SEPARABLE_COMPILATION_OBJECTS should be called.
#
#   CUDA_SOURCE_PROPERTY_FORMAT
#   -- If this source file property is set, it can override the format specified
#      to CUDA_WRAP_SRCS (OBJ, PTX, CUBIN, or FATBIN).  If an input source file
#      is not a .cu file, setting this file will cause it to be treated as a .cu
#      file. See documentation for set_source_files_properties on how to set
#      this property.
#
#   CUDA_USE_STATIC_CUDA_RUNTIME (Default ON)
#   -- When enabled the static version of the CUDA runtime library will be used
#      in CUDA_LIBRARIES.  If the version of CUDA configured doesn't support
#      this option, then it will be silently disabled.
#
#   CUDA_VERBOSE_BUILD (Default OFF)
#   -- Set to ON to see all the commands used when building the CUDA file.  When
#      using a Makefile generator the value defaults to VERBOSE (run make
#      VERBOSE=1 to see output), although setting CUDA_VERBOSE_BUILD to ON will
#      always print the output.
#
# The script creates the following macros (in alphebetical order)::
#
#   CUDA_ADD_CUFFT_TO_TARGET( cuda_target )
#   -- Adds the cufft library to the target (can be any target).  Handles whether
#      you are in emulation mode or not.
#
#   CUDA_ADD_CUBLAS_TO_TARGET( cuda_target )
#   -- Adds the cublas library to the target (can be any target).  Handles
#      whether you are in emulation mode or not.
#
#   CUDA_ADD_EXECUTABLE( cuda_target file0 file1 ...
#                        [WIN32] [MACOSX_BUNDLE] [EXCLUDE_FROM_ALL] [OPTIONS ...] )
#   -- Creates an executable "cuda_target" which is made up of the files
#      specified.  All of the non CUDA C files are compiled using the standard
#      build rules specified by CMAKE and the cuda files are compiled to object
#      files using nvcc and the host compiler.  In addition CUDA_INCLUDE_DIRS is
#      added automatically to include_directories().  Some standard CMake target
#      calls can be used on the target after calling this macro
#      (e.g. set_target_properties and target_link_libraries), but setting
#      properties that adjust compilation flags will not affect code compiled by
#      nvcc.  Such flags should be modified before calling CUDA_ADD_EXECUTABLE,
#      CUDA_ADD_LIBRARY or CUDA_WRAP_SRCS.
#
#   CUDA_ADD_LIBRARY( cuda_target file0 file1 ...
#                     [STATIC | SHARED | MODULE] [EXCLUDE_FROM_ALL] [OPTIONS ...] )
#   -- Same as CUDA_ADD_EXECUTABLE except that a library is created.
#
#   CUDA_BUILD_CLEAN_TARGET()
#   -- Creates a convience target that deletes all the dependency files
#      generated.  You should make clean after running this target to ensure the
#      dependency files get regenerated.
#
#   CUDA_COMPILE( generated_files file0 file1 ... [STATIC | SHARED | MODULE]
#                 [OPTIONS ...] )
#   -- Returns a list of generated files from the input source files to be used
#      with ADD_LIBRARY or ADD_EXECUTABLE.
#
#   CUDA_COMPILE_PTX( generated_files file0 file1 ... [OPTIONS ...] )
#   -- Returns a list of PTX files generated from the input source files.
#
#   CUDA_COMPILE_FATBIN( generated_files file0 file1 ... [OPTIONS ...] )
#   -- Returns a list of FATBIN files generated from the input source files.
#
#   CUDA_COMPILE_CUBIN( generated_files file0 file1 ... [OPTIONS ...] )
#   -- Returns a list of CUBIN files generated from the input source files.
#
#   CUDA_COMPUTE_SEPARABLE_COMPILATION_OBJECT_FILE_NAME( output_file_var
#                                                        cuda_target
#                                                        object_files )
#   -- Compute the name of the intermediate link file used for separable
#      compilation.  This file name is typically passed into
#      CUDA_LINK_SEPARABLE_COMPILATION_OBJECTS.  output_file_var is produced
#      based on cuda_target the list of objects files that need separable
#      compilation as specified by object_files.  If the object_files list is
#      empty, then output_file_var will be empty.  This function is called
#      automatically for CUDA_ADD_LIBRARY and CUDA_ADD_EXECUTABLE.  Note that
#      this is a function and not a macro.
#
#   CUDA_INCLUDE_DIRECTORIES( path0 path1 ... )
#   -- Sets the directories that should be passed to nvcc
#      (e.g. nvcc -Ipath0 -Ipath1 ... ). These paths usually contain other .cu
#      files.
#
#
#   CUDA_LINK_SEPARABLE_COMPILATION_OBJECTS( output_file_var cuda_target
#                                            nvcc_flags object_files)
#   -- Generates the link object required by separable compilation from the given
#      object files.  This is called automatically for CUDA_ADD_EXECUTABLE and
#      CUDA_ADD_LIBRARY, but can be called manually when using CUDA_WRAP_SRCS
#      directly.  When called from CUDA_ADD_LIBRARY or CUDA_ADD_EXECUTABLE the
#      nvcc_flags passed in are the same as the flags passed in via the OPTIONS
#      argument.  The only nvcc flag added automatically is the bitness flag as
#      specified by CUDA_64_BIT_DEVICE_CODE.  Note that this is a function
#      instead of a macro.
#
#   CUDA_SELECT_NVCC_ARCH_FLAGS(out_variable [target_CUDA_architectures])
#   -- Selects GPU arch flags for nvcc based on target_CUDA_architectures
#      target_CUDA_architectures : Auto | Common | All | LIST(ARCH_AND_PTX ...)
#       - "Auto" detects local machine GPU compute arch at runtime.
#       - "Common" and "All" cover common and entire subsets of architectures
#      ARCH_AND_PTX : NAME | NUM.NUM | NUM.NUM(NUM.NUM) | NUM.NUM+PTX
#      NAME: Fermi Kepler Maxwell Kepler+Tegra Kepler+Tesla Maxwell+Tegra Pascal
#      NUM: Any number. Only those pairs are currently accepted by NVCC though:
#            2.0 2.1 3.0 3.2 3.5 3.7 5.0 5.2 5.3 6.0 6.2
#      Returns LIST of flags to be added to CUDA_NVCC_FLAGS in ${out_variable}
#      Additionally, sets ${out_variable}_readable to the resulting numeric list
#      Example:
#       CUDA_SELECT_NVCC_ARCH_FLAGS(ARCH_FLAGS 3.0 3.5+PTX 5.2(5.0) Maxwell)
#        LIST(APPEND CUDA_NVCC_FLAGS ${ARCH_FLAGS})
#
#      More info on CUDA architectures: https://en.wikipedia.org/wiki/CUDA
#      Note that this is a function instead of a macro.
#
#   CUDA_WRAP_SRCS ( cuda_target format generated_files file0 file1 ...
#                    [STATIC | SHARED | MODULE] [OPTIONS ...] )
#   -- This is where all the magic happens.  CUDA_ADD_EXECUTABLE,
#      CUDA_ADD_LIBRARY, CUDA_COMPILE, and CUDA_COMPILE_PTX all call this
#      function under the hood.
#
#      Given the list of files (file0 file1 ... fileN) this macro generates
#      custom commands that generate either PTX or linkable objects (use "PTX" or
#      "OBJ" for the format argument to switch).  Files that don't end with .cu
#      or have the HEADER_FILE_ONLY property are ignored.
#
#      The arguments passed in after OPTIONS are extra command line options to
#      give to nvcc.  You can also specify per configuration options by
#      specifying the name of the configuration followed by the options.  General
#      options must precede configuration specific options.  Not all
#      configurations need to be specified, only the ones provided will be used.
#
#         OPTIONS -DFLAG=2 "-DFLAG_OTHER=space in flag"
#         DEBUG -g
#         RELEASE --use_fast_math
#         RELWITHDEBINFO --use_fast_math;-g
#         MINSIZEREL --use_fast_math
#
#      For certain configurations (namely VS generating object files with
#      CUDA_ATTACH_VS_BUILD_RULE_TO_CUDA_FILE set to ON), no generated file will
#      be produced for the given cuda file.  This is because when you add the
#      cuda file to Visual Studio it knows that this file produces an object file
#      and will link in the resulting object file automatically.
#
#      This script will also generate a separate cmake script that is used at
#      build time to invoke nvcc.  This is for several reasons.
#
#        1. nvcc can return negative numbers as return values which confuses
#        Visual Studio into thinking that the command succeeded.  The script now
#        checks the error codes and produces errors when there was a problem.
#
#        2. nvcc has been known to not delete incomplete results when it
#        encounters problems.  This confuses build systems into thinking the
#        target was generated when in fact an unusable file exists.  The script
#        now deletes the output files if there was an error.
#
#        3. By putting all the options that affect the build into a file and then
#        make the build rule dependent on the file, the output files will be
#        regenerated when the options change.
#
#      This script also looks at optional arguments STATIC, SHARED, or MODULE to
#      determine when to target the object compilation for a shared library.
#      BUILD_SHARED_LIBS is ignored in CUDA_WRAP_SRCS, but it is respected in
#      CUDA_ADD_LIBRARY.  On some systems special flags are added for building
#      objects intended for shared libraries.  A preprocessor macro,
#      <target_name>_EXPORTS is defined when a shared library compilation is
#      detected.
#
#      Flags passed into add_definitions with -D or /D are passed along to nvcc.
#
#
#
# The script defines the following variables::
#
#   CUDA_VERSION_MAJOR    -- The major version of cuda as reported by nvcc.
#   CUDA_VERSION_MINOR    -- The minor version.
#   CUDA_VERSION
#   CUDA_VERSION_STRING   -- CUDA_VERSION_MAJOR.CUDA_VERSION_MINOR
#   CUDA_HAS_FP16         -- Whether a short float (float16,fp16) is supported.
#
#   CUDA_TOOLKIT_ROOT_DIR -- Path to the CUDA Toolkit (defined if not set).
#   CUDA_SDK_ROOT_DIR     -- Path to the CUDA SDK.  Use this to find files in the
#                            SDK.  This script will not directly support finding
#                            specific libraries or headers, as that isn't
#                            supported by NVIDIA.  If you want to change
#                            libraries when the path changes see the
#                            FindCUDA.cmake script for an example of how to clear
#                            these variables.  There are also examples of how to
#                            use the CUDA_SDK_ROOT_DIR to locate headers or
#                            libraries, if you so choose (at your own risk).
#   CUDA_INCLUDE_DIRS     -- Include directory for cuda headers.  Added automatically
#                            for CUDA_ADD_EXECUTABLE and CUDA_ADD_LIBRARY.
#   CUDA_LIBRARIES        -- Cuda RT library.
#   CUDA_CUFFT_LIBRARIES  -- Device or emulation library for the Cuda FFT
#                            implementation (alternative to:
#                            CUDA_ADD_CUFFT_TO_TARGET macro)
#   CUDA_CUBLAS_LIBRARIES -- Device or emulation library for the Cuda BLAS
#                            implementation (alterative to:
#                            CUDA_ADD_CUBLAS_TO_TARGET macro).
#   CUDA_cudart_static_LIBRARY -- Statically linkable cuda runtime library.
#                                 Only available for CUDA version 5.5+
#   CUDA_cupti_LIBRARY    -- CUDA Profiling Tools Interface library.
#                            Only available for CUDA version 4.0+.
#   CUDA_curand_LIBRARY   -- CUDA Random Number Generation library.
#                            Only available for CUDA version 3.2+.
#   CUDA_cusolver_LIBRARY -- CUDA Direct Solver library.
#                            Only available for CUDA version 7.0+.
#   CUDA_cusparse_LIBRARY -- CUDA Sparse Matrix library.
#                            Only available for CUDA version 3.2+.
#   CUDA_npp_LIBRARY      -- NVIDIA Performance Primitives lib.
#                            Only available for CUDA version 4.0+.
#   CUDA_nppc_LIBRARY     -- NVIDIA Performance Primitives lib (core).
#                            Only available for CUDA version 5.5+.
#   CUDA_nppi_LIBRARY     -- NVIDIA Performance Primitives lib (image processing).
#                            Only available for CUDA version 5.5+.
#   CUDA_npps_LIBRARY     -- NVIDIA Performance Primitives lib (signal processing).
#                            Only available for CUDA version 5.5+.
#   CUDA_nvcuvenc_LIBRARY -- CUDA Video Encoder library.
#                            Only available for CUDA version 3.2+.
#                            Windows only.
#   CUDA_nvcuvid_LIBRARY  -- CUDA Video Decoder library.
#                            Only available for CUDA version 3.2+.
#                            Windows only.
#

#   James Bigler, NVIDIA Corp (nvidia.com - jbigler)
#   Abe Stephens, SCI Institute -- http://www.sci.utah.edu/~abe/FindCuda.html
#
#   Copyright (c) 2008 - 2009 NVIDIA Corporation.  All rights reserved.
#
#   Copyright (c) 2007-2009
#   Scientific Computing and Imaging Institute, University of Utah
#
#   This code is licensed under the MIT License.  See the FindCUDA.cmake script
#   for the text of the license.

# The MIT License
#
# License for the specific language governing rights and limitations under
# Permission is hereby granted, free of charge, to any person obtaining a
# copy of this software and associated documentation files (the "Software"),
# to deal in the Software without restriction, including without limitation
# the rights to use, copy, modify, merge, publish, distribute, sublicense,
# and/or sell copies of the Software, and to permit persons to whom the
# Software is furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included
# in all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
# DEALINGS IN THE SOFTWARE.
#
###############################################################################

# FindCUDA.cmake

# We need to have at least this version to support the VERSION_LESS argument to 'if' (2.6.2) and unset (2.6.3)
cmake_policy(PUSH)
cmake_minimum_required(VERSION 2.6.3)
cmake_policy(POP)

# This macro helps us find the location of helper files we will need the full path to
macro(CUDA_FIND_HELPER_FILE _name _extension)
  set(_full_name "${_name}.${_extension}")
  # CMAKE_CURRENT_LIST_FILE contains the full path to the file currently being
  # processed.  Using this variable, we can pull out the current path, and
  # provide a way to get access to the other files we need local to here.
  get_filename_component(CMAKE_CURRENT_LIST_DIR "${CMAKE_CURRENT_LIST_FILE}" PATH)
  set(CUDA_${_name} "${CMAKE_CURRENT_LIST_DIR}/FindCUDA/${_full_name}")
  if(NOT EXISTS "${CUDA_${_name}}")
    set(error_message "${_full_name} not found in ${CMAKE_CURRENT_LIST_DIR}/FindCUDA")
    if(CUDA_FIND_REQUIRED)
      message(FATAL_ERROR "${error_message}")
    else()
      if(NOT CUDA_FIND_QUIETLY)
        message(STATUS "${error_message}")
      endif()
    endif()
  endif()
  # Set this variable as internal, so the user isn't bugged with it.
  set(CUDA_${_name} ${CUDA_${_name}} CACHE INTERNAL "Location of ${_full_name}" FORCE)
endmacro()

#####################################################################
## CUDA_INCLUDE_NVCC_DEPENDENCIES
##

# So we want to try and include the dependency file if it exists.  If
# it doesn't exist then we need to create an empty one, so we can
# include it.

# If it does exist, then we need to check to see if all the files it
# depends on exist.  If they don't then we should clear the dependency
# file and regenerate it later.  This covers the case where a header
# file has disappeared or moved.

macro(CUDA_INCLUDE_NVCC_DEPENDENCIES dependency_file)
  set(CUDA_NVCC_DEPEND)
  set(CUDA_NVCC_DEPEND_REGENERATE FALSE)


  # Include the dependency file.  Create it first if it doesn't exist .  The
  # INCLUDE puts a dependency that will force CMake to rerun and bring in the
  # new info when it changes.  DO NOT REMOVE THIS (as I did and spent a few
  # hours figuring out why it didn't work.
  if(NOT EXISTS ${dependency_file})
    file(WRITE ${dependency_file} "#FindCUDA.cmake generated file.  Do not edit.\n")
  endif()
  # Always include this file to force CMake to run again next
  # invocation and rebuild the dependencies.
  #message("including dependency_file = ${dependency_file}")
  include(${dependency_file})

  # Now we need to verify the existence of all the included files
  # here.  If they aren't there we need to just blank this variable and
  # make the file regenerate again.
#   if(DEFINED CUDA_NVCC_DEPEND)
#     message("CUDA_NVCC_DEPEND set")
#   else()
#     message("CUDA_NVCC_DEPEND NOT set")
#   endif()
  if(CUDA_NVCC_DEPEND)
    #message("CUDA_NVCC_DEPEND found")
    foreach(f ${CUDA_NVCC_DEPEND})
      # message("searching for ${f}")
      if(NOT EXISTS ${f})
        #message("file ${f} not found")
        set(CUDA_NVCC_DEPEND_REGENERATE TRUE)
      endif()
    endforeach()
  else()
    #message("CUDA_NVCC_DEPEND false")
    # No dependencies, so regenerate the file.
    set(CUDA_NVCC_DEPEND_REGENERATE TRUE)
  endif()

  #message("CUDA_NVCC_DEPEND_REGENERATE = ${CUDA_NVCC_DEPEND_REGENERATE}")
  # No incoming dependencies, so we need to generate them.  Make the
  # output depend on the dependency file itself, which should cause the
  # rule to re-run.
  if(CUDA_NVCC_DEPEND_REGENERATE)
    set(CUDA_NVCC_DEPEND ${dependency_file})
    #message("Generating an empty dependency_file: ${dependency_file}")
    file(WRITE ${dependency_file} "#FindCUDA.cmake generated file.  Do not edit.\n")
  endif()

endmacro()

###############################################################################
###############################################################################
# Setup variables' defaults
###############################################################################
###############################################################################

# Allow the user to specify if the device code is supposed to be 32 or 64 bit.
if(CMAKE_SIZEOF_VOID_P EQUAL 8)
  set(CUDA_64_BIT_DEVICE_CODE_DEFAULT ON)
else()
  set(CUDA_64_BIT_DEVICE_CODE_DEFAULT OFF)
endif()
option(CUDA_64_BIT_DEVICE_CODE "Compile device code in 64 bit mode" ${CUDA_64_BIT_DEVICE_CODE_DEFAULT})

# Attach the build rule to the source file in VS.  This option
option(CUDA_ATTACH_VS_BUILD_RULE_TO_CUDA_FILE "Attach the build rule to the CUDA source file.  Enable only when the CUDA source file is added to at most one target." ON)

# Prints out extra information about the cuda file during compilation
option(CUDA_BUILD_CUBIN "Generate and parse .cubin files in Device mode." OFF)

# Set whether we are using emulation or device mode.
option(CUDA_BUILD_EMULATION "Build in Emulation mode" OFF)

# Where to put the generated output.
set(CUDA_GENERATED_OUTPUT_DIR "" CACHE PATH "Directory to put all the output files.  If blank it will default to the CMAKE_CURRENT_BINARY_DIR")

# Parse HOST_COMPILATION mode.
option(CUDA_HOST_COMPILATION_CPP "Generated file extension" ON)

# Extra user settable flags
set(CUDA_NVCC_FLAGS "" CACHE STRING "Semi-colon delimit multiple arguments.")

if(CMAKE_GENERATOR MATCHES "Visual Studio")
  set(CUDA_HOST_COMPILER "$(VCInstallDir)bin" CACHE FILEPATH "Host side compiler used by NVCC")
else()
  if(APPLE
      AND "${CMAKE_C_COMPILER_ID}" MATCHES "Clang"
      AND "${CMAKE_C_COMPILER}" MATCHES "/cc$")
    # Using cc which is symlink to clang may let NVCC think it is GCC and issue
    # unhandled -dumpspecs option to clang. Also in case neither
    # CMAKE_C_COMPILER is defined (project does not use C language) nor
    # CUDA_HOST_COMPILER is specified manually we should skip -ccbin and let
    # nvcc use its own default C compiler.
    # Only care about this on APPLE with clang to avoid
    # following symlinks to things like ccache
    if(DEFINED CMAKE_C_COMPILER AND NOT DEFINED CUDA_HOST_COMPILER)
      get_filename_component(c_compiler_realpath "${CMAKE_C_COMPILER}" REALPATH)
      # if the real path does not end up being clang then
      # go back to using CMAKE_C_COMPILER
      if(NOT "${c_compiler_realpath}" MATCHES "/clang$")
        set(c_compiler_realpath "${CMAKE_C_COMPILER}")
      endif()
    else()
      set(c_compiler_realpath "")
    endif()
    set(CUDA_HOST_COMPILER "${c_compiler_realpath}" CACHE FILEPATH "Host side compiler used by NVCC")
  else()
    set(CUDA_HOST_COMPILER "${CMAKE_C_COMPILER}"
      CACHE FILEPATH "Host side compiler used by NVCC")
  endif()
endif()

# Propagate the host flags to the host compiler via -Xcompiler
option(CUDA_PROPAGATE_HOST_FLAGS "Propage C/CXX_FLAGS and friends to the host compiler via -Xcompile" ON)

# Enable CUDA_SEPARABLE_COMPILATION
option(CUDA_SEPARABLE_COMPILATION "Compile CUDA objects with separable compilation enabled.  Requires CUDA 5.0+" OFF)

# Specifies whether the commands used when compiling the .cu file will be printed out.
option(CUDA_VERBOSE_BUILD "Print out the commands run while compiling the CUDA source file.  With the Makefile generator this defaults to VERBOSE variable specified on the command line, but can be forced on with this option." OFF)

mark_as_advanced(
  CUDA_64_BIT_DEVICE_CODE
  CUDA_ATTACH_VS_BUILD_RULE_TO_CUDA_FILE
  CUDA_GENERATED_OUTPUT_DIR
  CUDA_HOST_COMPILATION_CPP
  CUDA_NVCC_FLAGS
  CUDA_PROPAGATE_HOST_FLAGS
  CUDA_BUILD_CUBIN
  CUDA_BUILD_EMULATION
  CUDA_VERBOSE_BUILD
  CUDA_SEPARABLE_COMPILATION
  )

# Makefile and similar generators don't define CMAKE_CONFIGURATION_TYPES, so we
# need to add another entry for the CMAKE_BUILD_TYPE.  We also need to add the
# standerd set of 4 build types (Debug, MinSizeRel, Release, and RelWithDebInfo)
# for completeness.  We need run this loop in order to accomodate the addition
# of extra configuration types.  Duplicate entries will be removed by
# REMOVE_DUPLICATES.
set(CUDA_configuration_types ${CMAKE_CONFIGURATION_TYPES} ${CMAKE_BUILD_TYPE} Debug MinSizeRel Release RelWithDebInfo)
list(REMOVE_DUPLICATES CUDA_configuration_types)
foreach(config ${CUDA_configuration_types})
    string(TOUPPER ${config} config_upper)
    set(CUDA_NVCC_FLAGS_${config_upper} "" CACHE STRING "Semi-colon delimit multiple arguments.")
    mark_as_advanced(CUDA_NVCC_FLAGS_${config_upper})
endforeach()

###############################################################################
###############################################################################
# Locate CUDA, Set Build Type, etc.
###############################################################################
###############################################################################

macro(cuda_unset_include_and_libraries)
  unset(CUDA_TOOLKIT_INCLUDE CACHE)
  unset(CUDA_CUDART_LIBRARY CACHE)
  unset(CUDA_CUDA_LIBRARY CACHE)
  # Make sure you run this before you unset CUDA_VERSION.
  if(CUDA_VERSION VERSION_EQUAL "3.0")
    # This only existed in the 3.0 version of the CUDA toolkit
    unset(CUDA_CUDARTEMU_LIBRARY CACHE)
  endif()
  unset(CUDA_cudart_static_LIBRARY CACHE)
  unset(CUDA_cublas_LIBRARY CACHE)
  unset(CUDA_cublas_device_LIBRARY CACHE)
  unset(CUDA_cublasemu_LIBRARY CACHE)
  unset(CUDA_cufft_LIBRARY CACHE)
  unset(CUDA_cufftemu_LIBRARY CACHE)
  unset(CUDA_cupti_LIBRARY CACHE)
  unset(CUDA_curand_LIBRARY CACHE)
  unset(CUDA_cusolver_LIBRARY CACHE)
  unset(CUDA_cusparse_LIBRARY CACHE)
  unset(CUDA_npp_LIBRARY CACHE)
  unset(CUDA_nppc_LIBRARY CACHE)
  unset(CUDA_nppi_LIBRARY CACHE)
  unset(CUDA_npps_LIBRARY CACHE)
  unset(CUDA_nvcuvenc_LIBRARY CACHE)
  unset(CUDA_nvcuvid_LIBRARY CACHE)
  unset(CUDA_USE_STATIC_CUDA_RUNTIME CACHE)
  unset(CUDA_GPU_DETECT_OUTPUT CACHE)
endmacro()

# Check to see if the CUDA_TOOLKIT_ROOT_DIR and CUDA_SDK_ROOT_DIR have changed,
# if they have then clear the cache variables, so that will be detected again.
if(NOT "${CUDA_TOOLKIT_ROOT_DIR}" STREQUAL "${CUDA_TOOLKIT_ROOT_DIR_INTERNAL}")
  unset(CUDA_TOOLKIT_TARGET_DIR CACHE)
  unset(CUDA_NVCC_EXECUTABLE CACHE)
  cuda_unset_include_and_libraries()
  unset(CUDA_VERSION CACHE)
endif()

if(NOT "${CUDA_TOOLKIT_TARGET_DIR}" STREQUAL "${CUDA_TOOLKIT_TARGET_DIR_INTERNAL}")
  cuda_unset_include_and_libraries()
endif()

#
#  End of unset()
#

#
#  Start looking for things
#

# Search for the cuda distribution.
if(NOT CUDA_TOOLKIT_ROOT_DIR AND NOT CMAKE_CROSSCOMPILING)
  # Search in the CUDA_BIN_PATH first.
  find_path(CUDA_TOOLKIT_ROOT_DIR
    NAMES nvcc nvcc.exe
    PATHS
      ENV CUDA_TOOLKIT_ROOT
      ENV CUDA_PATH
      ENV CUDA_BIN_PATH
    PATH_SUFFIXES bin bin64
    DOC "Toolkit location."
    NO_DEFAULT_PATH
    )

  # Now search default paths
  find_path(CUDA_TOOLKIT_ROOT_DIR
    NAMES nvcc nvcc.exe
    PATHS /usr/local/bin
          /usr/local/cuda/bin
    DOC "Toolkit location."
    )

  if (CUDA_TOOLKIT_ROOT_DIR)
    string(REGEX REPLACE "[/\\\\]?bin[64]*[/\\\\]?$" "" CUDA_TOOLKIT_ROOT_DIR ${CUDA_TOOLKIT_ROOT_DIR})
    # We need to force this back into the cache.
    set(CUDA_TOOLKIT_ROOT_DIR ${CUDA_TOOLKIT_ROOT_DIR} CACHE PATH "Toolkit location." FORCE)
    set(CUDA_TOOLKIT_TARGET_DIR ${CUDA_TOOLKIT_ROOT_DIR})
  endif()

  if (NOT EXISTS ${CUDA_TOOLKIT_ROOT_DIR})
    if(CUDA_FIND_REQUIRED)
      message(FATAL_ERROR "Specify CUDA_TOOLKIT_ROOT_DIR")
    elseif(NOT CUDA_FIND_QUIETLY)
      message("CUDA_TOOLKIT_ROOT_DIR not found or specified")
    endif()
  endif ()
endif ()

if(CMAKE_CROSSCOMPILING)
  SET (CUDA_TOOLKIT_ROOT $ENV{CUDA_TOOLKIT_ROOT})
  if(CMAKE_SYSTEM_PROCESSOR STREQUAL "armv7-a")
    # Support for NVPACK
    set (CUDA_TOOLKIT_TARGET_NAME "armv7-linux-androideabi")
  elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "arm")
    # Support for arm cross compilation
    set(CUDA_TOOLKIT_TARGET_NAME "armv7-linux-gnueabihf")
  elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64")
    # Support for aarch64 cross compilation
    if (ANDROID_ARCH_NAME STREQUAL "arm64")
      set(CUDA_TOOLKIT_TARGET_NAME "aarch64-linux-androideabi")
    else()
      set(CUDA_TOOLKIT_TARGET_NAME "aarch64-linux")
    endif (ANDROID_ARCH_NAME STREQUAL "arm64")
  endif()

  if (EXISTS "${CUDA_TOOLKIT_ROOT}/targets/${CUDA_TOOLKIT_TARGET_NAME}")
    set(CUDA_TOOLKIT_TARGET_DIR "${CUDA_TOOLKIT_ROOT}/targets/${CUDA_TOOLKIT_TARGET_NAME}" CACHE PATH "CUDA Toolkit target location.")
    SET (CUDA_TOOLKIT_ROOT_DIR ${CUDA_TOOLKIT_ROOT})
    mark_as_advanced(CUDA_TOOLKIT_TARGET_DIR)
  endif()

  # add known CUDA targetr root path to the set of directories we search for programs, libraries and headers
  set( CMAKE_FIND_ROOT_PATH "${CUDA_TOOLKIT_TARGET_DIR};${CMAKE_FIND_ROOT_PATH}")
  macro( cuda_find_host_program )
    find_host_program( ${ARGN} )
  endmacro()
else()
  # for non-cross-compile, find_host_program == find_program and CUDA_TOOLKIT_TARGET_DIR == CUDA_TOOLKIT_ROOT_DIR
  macro( cuda_find_host_program )
    find_program( ${ARGN} )
  endmacro()
  SET (CUDA_TOOLKIT_TARGET_DIR ${CUDA_TOOLKIT_ROOT_DIR})
endif()


# CUDA_NVCC_EXECUTABLE
cuda_find_host_program(CUDA_NVCC_EXECUTABLE
  NAMES nvcc
  PATHS "${CUDA_TOOLKIT_ROOT_DIR}"
  ENV CUDA_PATH
  ENV CUDA_BIN_PATH
  PATH_SUFFIXES bin bin64
  NO_DEFAULT_PATH
  )
# Search default search paths, after we search our own set of paths.
cuda_find_host_program(CUDA_NVCC_EXECUTABLE nvcc)
mark_as_advanced(CUDA_NVCC_EXECUTABLE)

if(CUDA_NVCC_EXECUTABLE AND NOT CUDA_VERSION)
  # Compute the version.
  execute_process (COMMAND ${CUDA_NVCC_EXECUTABLE} "--version" OUTPUT_VARIABLE NVCC_OUT)
  string(REGEX REPLACE ".*release ([0-9]+)\\.([0-9]+).*" "\\1" CUDA_VERSION_MAJOR ${NVCC_OUT})
  string(REGEX REPLACE ".*release ([0-9]+)\\.([0-9]+).*" "\\2" CUDA_VERSION_MINOR ${NVCC_OUT})
  set(CUDA_VERSION "${CUDA_VERSION_MAJOR}.${CUDA_VERSION_MINOR}" CACHE STRING "Version of CUDA as computed from nvcc.")
  mark_as_advanced(CUDA_VERSION)
else()
  # Need to set these based off of the cached value
  string(REGEX REPLACE "([0-9]+)\\.([0-9]+).*" "\\1" CUDA_VERSION_MAJOR "${CUDA_VERSION}")
  string(REGEX REPLACE "([0-9]+)\\.([0-9]+).*" "\\2" CUDA_VERSION_MINOR "${CUDA_VERSION}")
endif()


# Always set this convenience variable
set(CUDA_VERSION_STRING "${CUDA_VERSION}")

# CUDA_TOOLKIT_INCLUDE
find_path(CUDA_TOOLKIT_INCLUDE
  device_functions.h # Header included in toolkit
  PATHS ${CUDA_TOOLKIT_TARGET_DIR}
  ENV CUDA_PATH
  ENV CUDA_INC_PATH
  PATH_SUFFIXES include
  NO_DEFAULT_PATH
  )
# Search default search paths, after we search our own set of paths.
find_path(CUDA_TOOLKIT_INCLUDE device_functions.h)
mark_as_advanced(CUDA_TOOLKIT_INCLUDE)

if (CUDA_VERSION VERSION_GREATER "7.0" OR EXISTS "${CUDA_TOOLKIT_INCLUDE}/cuda_fp16.h")
  set(CUDA_HAS_FP16 TRUE)
else()
  set(CUDA_HAS_FP16 FALSE)
endif()

# Set the user list of include dir to nothing to initialize it.
set (CUDA_NVCC_INCLUDE_ARGS_USER "")
set (CUDA_INCLUDE_DIRS ${CUDA_TOOLKIT_INCLUDE})

macro(cuda_find_library_local_first_with_path_ext _var _names _doc _path_ext )
  if(CMAKE_SIZEOF_VOID_P EQUAL 8)
    # CUDA 3.2+ on Windows moved the library directories, so we need the new
    # and old paths.
    set(_cuda_64bit_lib_dir "${_path_ext}lib/x64" "${_path_ext}lib64" "${_path_ext}libx64" )
  endif()
  # CUDA 3.2+ on Windows moved the library directories, so we need to new
  # (lib/Win32) and the old path (lib).
  find_library(${_var}
    NAMES ${_names}
    PATHS "${CUDA_TOOLKIT_TARGET_DIR}"
    ENV CUDA_PATH
    ENV CUDA_LIB_PATH
    PATH_SUFFIXES ${_cuda_64bit_lib_dir} "${_path_ext}lib/Win32" "${_path_ext}lib" "${_path_ext}libWin32"
    DOC ${_doc}
    NO_DEFAULT_PATH
    )
  if (NOT CMAKE_CROSSCOMPILING)
    # Search default search paths, after we search our own set of paths.
    find_library(${_var}
      NAMES ${_names}
      PATHS "/usr/lib/nvidia-current"
      DOC ${_doc}
      )
  endif()
endmacro()

macro(cuda_find_library_local_first _var _names _doc)
  cuda_find_library_local_first_with_path_ext( "${_var}" "${_names}" "${_doc}" "" )
endmacro()

macro(find_library_local_first _var _names _doc )
  cuda_find_library_local_first( "${_var}" "${_names}" "${_doc}" "" )
endmacro()


# CUDA_LIBRARIES
cuda_find_library_local_first(CUDA_CUDART_LIBRARY cudart "\"cudart\" library")
if(CUDA_VERSION VERSION_EQUAL "3.0")
  # The cudartemu library only existed for the 3.0 version of CUDA.
  cuda_find_library_local_first(CUDA_CUDARTEMU_LIBRARY cudartemu "\"cudartemu\" library")
  mark_as_advanced(
    CUDA_CUDARTEMU_LIBRARY
    )
endif()

if(CUDA_USE_STATIC_CUDA_RUNTIME AND NOT CUDA_VERSION VERSION_LESS "5.5")
  cuda_find_library_local_first(CUDA_cudart_static_LIBRARY cudart_static "static CUDA runtime library")
  mark_as_advanced(CUDA_cudart_static_LIBRARY)
endif()


if(CUDA_cudart_static_LIBRARY)
  # Set whether to use the static cuda runtime.
  option(CUDA_USE_STATIC_CUDA_RUNTIME "Use the static version of the CUDA runtime library if available" ON)
  set(CUDA_CUDART_LIBRARY_VAR CUDA_cudart_static_LIBRARY)
else()
  option(CUDA_USE_STATIC_CUDA_RUNTIME "Use the static version of the CUDA runtime library if available" OFF)
  set(CUDA_CUDART_LIBRARY_VAR CUDA_CUDART_LIBRARY)
endif()

if(CUDA_USE_STATIC_CUDA_RUNTIME)
  if(UNIX)
    # Check for the dependent libraries.  Here we look for pthreads.
    if (DEFINED CMAKE_THREAD_PREFER_PTHREAD)
      set(_cuda_cmake_thread_prefer_pthread ${CMAKE_THREAD_PREFER_PTHREAD})
    endif()
    set(CMAKE_THREAD_PREFER_PTHREAD 1)

    # Many of the FindXYZ CMake comes with makes use of try_compile with int main(){return 0;}
    # as the source file.  Unfortunately this causes a warning with -Wstrict-prototypes and
    # -Werror causes the try_compile to fail.  We will just temporarily disable other flags
    # when doing the find_package command here.
    set(_cuda_cmake_c_flags ${CMAKE_C_FLAGS})
    set(CMAKE_C_FLAGS "-fPIC")
    find_package(Threads REQUIRED)
    set(CMAKE_C_FLAGS ${_cuda_cmake_c_flags})

    if (DEFINED _cuda_cmake_thread_prefer_pthread)
      set(CMAKE_THREAD_PREFER_PTHREAD ${_cuda_cmake_thread_prefer_pthread})
      unset(_cuda_cmake_thread_prefer_pthread)
    else()
      unset(CMAKE_THREAD_PREFER_PTHREAD)
    endif()
  endif()
  if (NOT APPLE AND CUDA_VERSION VERSION_LESS "7.0")
    # Before CUDA 7.0, there was librt that has things such as, clock_gettime, shm_open, and shm_unlink.
    find_library(CUDA_rt_LIBRARY rt)
    if (NOT CUDA_rt_LIBRARY)
      message(WARNING "Expecting to find librt for libcudart_static, but didn't find it.")
    endif()
  endif()
endif()

# CUPTI library showed up in cuda toolkit 4.0
if(NOT CUDA_VERSION VERSION_LESS "4.0")
  cuda_find_library_local_first_with_path_ext(CUDA_cupti_LIBRARY cupti "\"cupti\" library" "extras/CUPTI/")
  mark_as_advanced(CUDA_cupti_LIBRARY)
endif()

# Set the CUDA_LIBRARIES variable.  This is the set of stuff to link against if you are
# using the CUDA runtime.  For the dynamic version of the runtime, most of the
# dependencies are brough in, but for the static version there are additional libraries
# and linker commands needed.
# Initialize to empty
set(CUDA_LIBRARIES)

# If we are using emulation mode and we found the cudartemu library then use
# that one instead of cudart.
if(CUDA_BUILD_EMULATION AND CUDA_CUDARTEMU_LIBRARY)
  list(APPEND CUDA_LIBRARIES ${CUDA_CUDARTEMU_LIBRARY})
elseif(CUDA_USE_STATIC_CUDA_RUNTIME AND CUDA_cudart_static_LIBRARY)
  list(APPEND CUDA_LIBRARIES ${CUDA_cudart_static_LIBRARY} ${CMAKE_THREAD_LIBS_INIT} ${CMAKE_DL_LIBS})
  if (CUDA_rt_LIBRARY)
    list(APPEND CUDA_LIBRARIES ${CUDA_rt_LIBRARY})
  endif()
  if(APPLE)
    # We need to add the default path to the driver (libcuda.dylib) as an rpath, so that
    # the static cuda runtime can find it at runtime.
    list(APPEND CUDA_LIBRARIES -Wl,-rpath,/usr/local/cuda/lib)
  endif()
else()
  list(APPEND CUDA_LIBRARIES ${CUDA_CUDART_LIBRARY})
endif()

# 1.1 toolkit on linux doesn't appear to have a separate library on
# some platforms.
cuda_find_library_local_first(CUDA_CUDA_LIBRARY cuda "\"cuda\" library (older versions only).")

mark_as_advanced(
  CUDA_CUDA_LIBRARY
  CUDA_CUDART_LIBRARY
  )

#######################
# Look for some of the toolkit helper libraries
macro(FIND_CUDA_HELPER_LIBS _name)
  cuda_find_library_local_first(CUDA_${_name}_LIBRARY ${_name} "\"${_name}\" library")
  mark_as_advanced(CUDA_${_name}_LIBRARY)
endmacro()

#######################
# Disable emulation for v3.1 onward
if(CUDA_VERSION VERSION_GREATER "3.0")
  if(CUDA_BUILD_EMULATION)
    message(FATAL_ERROR "CUDA_BUILD_EMULATION is not supported in version 3.1 and onwards.  You must disable it to proceed.  You have version ${CUDA_VERSION}.")
  endif()
endif()

# Search for additional CUDA toolkit libraries.
if(CUDA_VERSION VERSION_LESS "3.1")
  # Emulation libraries aren't available in version 3.1 onward.
  find_cuda_helper_libs(cufftemu)
  find_cuda_helper_libs(cublasemu)
endif()
find_cuda_helper_libs(cufft)
find_cuda_helper_libs(cublas)
if(NOT CUDA_VERSION VERSION_LESS "3.2")
  # cusparse showed up in version 3.2
  find_cuda_helper_libs(cusparse)
  find_cuda_helper_libs(curand)
  if (WIN32)
    find_cuda_helper_libs(nvcuvenc)
    find_cuda_helper_libs(nvcuvid)
  endif()
endif()
if(CUDA_VERSION VERSION_GREATER "5.0")
  find_cuda_helper_libs(cublas_device)
  # In CUDA 5.5 NPP was splitted onto 3 separate libraries.
  find_cuda_helper_libs(nppc)
  find_cuda_helper_libs(nppi)
  find_cuda_helper_libs(npps)
  set(CUDA_npp_LIBRARY "${CUDA_nppc_LIBRARY};${CUDA_nppi_LIBRARY};${CUDA_npps_LIBRARY}")
elseif(NOT CUDA_VERSION VERSION_LESS "4.0")
  find_cuda_helper_libs(npp)
endif()
if(NOT CUDA_VERSION VERSION_LESS "7.0")
  # cusolver showed up in version 7.0
  find_cuda_helper_libs(cusolver)
endif()

if (CUDA_BUILD_EMULATION)
  set(CUDA_CUFFT_LIBRARIES ${CUDA_cufftemu_LIBRARY})
  set(CUDA_CUBLAS_LIBRARIES ${CUDA_cublasemu_LIBRARY})
else()
  set(CUDA_CUFFT_LIBRARIES ${CUDA_cufft_LIBRARY})
  set(CUDA_CUBLAS_LIBRARIES ${CUDA_cublas_LIBRARY} ${CUDA_cublas_device_LIBRARY})
endif()

########################
# Look for the SDK stuff.  As of CUDA 3.0 NVSDKCUDA_ROOT has been replaced with
# NVSDKCOMPUTE_ROOT with the old CUDA C contents moved into the C subdirectory
find_path(CUDA_SDK_ROOT_DIR common/inc/cutil.h
 HINTS
  "$ENV{NVSDKCOMPUTE_ROOT}/C"
  ENV NVSDKCUDA_ROOT
  "[HKEY_LOCAL_MACHINE\\SOFTWARE\\NVIDIA Corporation\\Installed Products\\NVIDIA SDK 10\\Compute;InstallDir]"
 PATHS
  "/Developer/GPU\ Computing/C"
  )

# Keep the CUDA_SDK_ROOT_DIR first in order to be able to override the
# environment variables.
set(CUDA_SDK_SEARCH_PATH
  "${CUDA_SDK_ROOT_DIR}"
  "${CUDA_TOOLKIT_ROOT_DIR}/local/NVSDK0.2"
  "${CUDA_TOOLKIT_ROOT_DIR}/NVSDK0.2"
  "${CUDA_TOOLKIT_ROOT_DIR}/NV_CUDA_SDK"
  "$ENV{HOME}/NVIDIA_CUDA_SDK"
  "$ENV{HOME}/NVIDIA_CUDA_SDK_MACOSX"
  "/Developer/CUDA"
  )

# Example of how to find an include file from the CUDA_SDK_ROOT_DIR

# find_path(CUDA_CUT_INCLUDE_DIR
#   cutil.h
#   PATHS ${CUDA_SDK_SEARCH_PATH}
#   PATH_SUFFIXES "common/inc"
#   DOC "Location of cutil.h"
#   NO_DEFAULT_PATH
#   )
# # Now search system paths
# find_path(CUDA_CUT_INCLUDE_DIR cutil.h DOC "Location of cutil.h")

# mark_as_advanced(CUDA_CUT_INCLUDE_DIR)


# Example of how to find a library in the CUDA_SDK_ROOT_DIR

# # cutil library is called cutil64 for 64 bit builds on windows.  We don't want
# # to get these confused, so we are setting the name based on the word size of
# # the build.

# if(CMAKE_SIZEOF_VOID_P EQUAL 8)
#   set(cuda_cutil_name cutil64)
# else()
#   set(cuda_cutil_name cutil32)
# endif()

# find_library(CUDA_CUT_LIBRARY
#   NAMES cutil ${cuda_cutil_name}
#   PATHS ${CUDA_SDK_SEARCH_PATH}
#   # The new version of the sdk shows up in common/lib, but the old one is in lib
#   PATH_SUFFIXES "common/lib" "lib"
#   DOC "Location of cutil library"
#   NO_DEFAULT_PATH
#   )
# # Now search system paths
# find_library(CUDA_CUT_LIBRARY NAMES cutil ${cuda_cutil_name} DOC "Location of cutil library")
# mark_as_advanced(CUDA_CUT_LIBRARY)
# set(CUDA_CUT_LIBRARIES ${CUDA_CUT_LIBRARY})



#############################
# Check for required components
set(CUDA_FOUND TRUE)

set(CUDA_TOOLKIT_ROOT_DIR_INTERNAL "${CUDA_TOOLKIT_ROOT_DIR}" CACHE INTERNAL
  "This is the value of the last time CUDA_TOOLKIT_ROOT_DIR was set successfully." FORCE)
set(CUDA_TOOLKIT_TARGET_DIR_INTERNAL "${CUDA_TOOLKIT_TARGET_DIR}" CACHE INTERNAL
  "This is the value of the last time CUDA_TOOLKIT_TARGET_DIR was set successfully." FORCE)
set(CUDA_SDK_ROOT_DIR_INTERNAL "${CUDA_SDK_ROOT_DIR}" CACHE INTERNAL
  "This is the value of the last time CUDA_SDK_ROOT_DIR was set successfully." FORCE)

#include(${CMAKE_CURRENT_LIST_DIR}/FindPackageHandleStandardArgs.cmake)
include(${CMAKE_ROOT}/Modules/FindPackageHandleStandardArgs.cmake)

find_package_handle_standard_args(CUDA
  REQUIRED_VARS
    CUDA_TOOLKIT_ROOT_DIR
    CUDA_NVCC_EXECUTABLE
    CUDA_INCLUDE_DIRS
    ${CUDA_CUDART_LIBRARY_VAR}
  VERSION_VAR
    CUDA_VERSION
  )



###############################################################################
###############################################################################
# Macros
###############################################################################
###############################################################################

###############################################################################
# Add include directories to pass to the nvcc command.
macro(CUDA_INCLUDE_DIRECTORIES)
  foreach(dir ${ARGN})
    list(APPEND CUDA_NVCC_INCLUDE_ARGS_USER -I${dir})
  endforeach()
endmacro()


##############################################################################
cuda_find_helper_file(parse_cubin cmake)
cuda_find_helper_file(make2cmake cmake)
cuda_find_helper_file(run_nvcc cmake)
include("${CMAKE_CURRENT_LIST_DIR}/FindCUDA/select_compute_arch.cmake")

##############################################################################
# Separate the OPTIONS out from the sources
#
macro(CUDA_GET_SOURCES_AND_OPTIONS _sources _cmake_options _options)
  set( ${_sources} )
  set( ${_cmake_options} )
  set( ${_options} )
  set( _found_options FALSE )
  foreach(arg ${ARGN})
    if("x${arg}" STREQUAL "xOPTIONS")
      set( _found_options TRUE )
    elseif(
        "x${arg}" STREQUAL "xWIN32" OR
        "x${arg}" STREQUAL "xMACOSX_BUNDLE" OR
        "x${arg}" STREQUAL "xEXCLUDE_FROM_ALL" OR
        "x${arg}" STREQUAL "xSTATIC" OR
        "x${arg}" STREQUAL "xSHARED" OR
        "x${arg}" STREQUAL "xMODULE"
        )
      list(APPEND ${_cmake_options} ${arg})
    else()
      if ( _found_options )
        list(APPEND ${_options} ${arg})
      else()
        # Assume this is a file
        list(APPEND ${_sources} ${arg})
      endif()
    endif()
  endforeach()
endmacro()

##############################################################################
# Parse the OPTIONS from ARGN and set the variables prefixed by _option_prefix
#
macro(CUDA_PARSE_NVCC_OPTIONS _option_prefix)
  set( _found_config )
  foreach(arg ${ARGN})
    # Determine if we are dealing with a perconfiguration flag
    foreach(config ${CUDA_configuration_types})
      string(TOUPPER ${config} config_upper)
      if (arg STREQUAL "${config_upper}")
        set( _found_config _${arg})
        # Set arg to nothing to keep it from being processed further
        set( arg )
      endif()
    endforeach()

    if ( arg )
      list(APPEND ${_option_prefix}${_found_config} "${arg}")
    endif()
  endforeach()
endmacro()

##############################################################################
# Helper to add the include directory for CUDA only once
function(CUDA_ADD_CUDA_INCLUDE_ONCE)
  get_directory_property(_include_directories INCLUDE_DIRECTORIES)
  set(_add TRUE)
  if(_include_directories)
    foreach(dir ${_include_directories})
      if("${dir}" STREQUAL "${CUDA_INCLUDE_DIRS}")
        set(_add FALSE)
      endif()
    endforeach()
  endif()
  if(_add)
    include_directories(${CUDA_INCLUDE_DIRS})
  endif()
endfunction()

function(CUDA_BUILD_SHARED_LIBRARY shared_flag)
  set(cmake_args ${ARGN})
  # If SHARED, MODULE, or STATIC aren't already in the list of arguments, then
  # add SHARED or STATIC based on the value of BUILD_SHARED_LIBS.
  list(FIND cmake_args SHARED _cuda_found_SHARED)
  list(FIND cmake_args MODULE _cuda_found_MODULE)
  list(FIND cmake_args STATIC _cuda_found_STATIC)
  if( _cuda_found_SHARED GREATER -1 OR
      _cuda_found_MODULE GREATER -1 OR
      _cuda_found_STATIC GREATER -1)
    set(_cuda_build_shared_libs)
  else()
    if (BUILD_SHARED_LIBS)
      set(_cuda_build_shared_libs SHARED)
    else()
      set(_cuda_build_shared_libs STATIC)
    endif()
  endif()
  set(${shared_flag} ${_cuda_build_shared_libs} PARENT_SCOPE)
endfunction()

##############################################################################
# Helper to avoid clashes of files with the same basename but different paths.
# This doesn't attempt to do exactly what CMake internals do, which is to only
# add this path when there is a conflict, since by the time a second collision
# in names is detected it's already too late to fix the first one.  For
# consistency sake the relative path will be added to all files.
function(CUDA_COMPUTE_BUILD_PATH path build_path)
  #message("CUDA_COMPUTE_BUILD_PATH([${path}] ${build_path})")
  # Only deal with CMake style paths from here on out
  file(TO_CMAKE_PATH "${path}" bpath)
  if (IS_ABSOLUTE "${bpath}")
    # Absolute paths are generally unnessary, especially if something like
    # file(GLOB_RECURSE) is used to pick up the files.

    string(FIND "${bpath}" "${CMAKE_CURRENT_BINARY_DIR}" _binary_dir_pos)
    if (_binary_dir_pos EQUAL 0)
      file(RELATIVE_PATH bpath "${CMAKE_CURRENT_BINARY_DIR}" "${bpath}")
    else()
      file(RELATIVE_PATH bpath "${CMAKE_CURRENT_SOURCE_DIR}" "${bpath}")
    endif()
  endif()

  # This recipe is from cmLocalGenerator::CreateSafeUniqueObjectFileName in the
  # CMake source.

  # Remove leading /
  string(REGEX REPLACE "^[/]+" "" bpath "${bpath}")
  # Avoid absolute paths by removing ':'
  string(REPLACE ":" "_" bpath "${bpath}")
  # Avoid relative paths that go up the tree
  string(REPLACE "../" "__/" bpath "${bpath}")
  # Avoid spaces
  string(REPLACE " " "_" bpath "${bpath}")

  # Strip off the filename.  I wait until here to do it, since removin the
  # basename can make a path that looked like path/../basename turn into
  # path/.. (notice the trailing slash).
  get_filename_component(bpath "${bpath}" PATH)

  set(${build_path} "${bpath}" PARENT_SCOPE)
  #message("${build_path} = ${bpath}")
endfunction()

##############################################################################
# This helper macro populates the following variables and setups up custom
# commands and targets to invoke the nvcc compiler to generate C or PTX source
# dependent upon the format parameter.  The compiler is invoked once with -M
# to generate a dependency file and a second time with -cuda or -ptx to generate
# a .cpp or .ptx file.
# INPUT:
#   cuda_target         - Target name
#   format              - PTX, CUBIN, FATBIN or OBJ
#   FILE1 .. FILEN      - The remaining arguments are the sources to be wrapped.
#   OPTIONS             - Extra options to NVCC
# OUTPUT:
#   generated_files     - List of generated files
##############################################################################
##############################################################################

macro(CUDA_WRAP_SRCS cuda_target format generated_files)

  # If CMake doesn't support separable compilation, complain
  if(CUDA_SEPARABLE_COMPILATION AND CMAKE_VERSION VERSION_LESS "2.8.10.1")
    message(SEND_ERROR "CUDA_SEPARABLE_COMPILATION isn't supported for CMake versions less than 2.8.10.1")
  endif()

  # Set up all the command line flags here, so that they can be overridden on a per target basis.

  set(nvcc_flags "")

  # Emulation if the card isn't present.
  if (CUDA_BUILD_EMULATION)
    # Emulation.
    set(nvcc_flags ${nvcc_flags} --device-emulation -D_DEVICEEMU -g)
  else()
    # Device mode.  No flags necessary.
  endif()

  if(CUDA_HOST_COMPILATION_CPP)
    set(CUDA_C_OR_CXX CXX)
  else()
    if(CUDA_VERSION VERSION_LESS "3.0")
      set(nvcc_flags ${nvcc_flags} --host-compilation C)
    else()
      message(WARNING "--host-compilation flag is deprecated in CUDA version >= 3.0.  Removing --host-compilation C flag" )
    endif()
    set(CUDA_C_OR_CXX C)
  endif()

  set(generated_extension ${CMAKE_${CUDA_C_OR_CXX}_OUTPUT_EXTENSION})

  if(CUDA_64_BIT_DEVICE_CODE)
    set(nvcc_flags ${nvcc_flags} -m64)
  else()
    set(nvcc_flags ${nvcc_flags} -m32)
  endif()

  if(CUDA_TARGET_CPU_ARCH)
    set(nvcc_flags ${nvcc_flags} "--target-cpu-architecture=${CUDA_TARGET_CPU_ARCH}")
  endif()

  # This needs to be passed in at this stage, because VS needs to fill out the
  # value of VCInstallDir from within VS.  Note that CCBIN is only used if
  # -ccbin or --compiler-bindir isn't used and CUDA_HOST_COMPILER matches
  # $(VCInstallDir)/bin.
  if(CMAKE_GENERATOR MATCHES "Visual Studio")
    set(ccbin_flags -D "\"CCBIN:PATH=$(VCInstallDir)bin\"" )
  else()
    set(ccbin_flags)
  endif()

  # Figure out which configure we will use and pass that in as an argument to
  # the script.  We need to defer the decision until compilation time, because
  # for VS projects we won't know if we are making a debug or release build
  # until build time.
  if(CMAKE_GENERATOR MATCHES "Visual Studio")
    set( CUDA_build_configuration "$(ConfigurationName)" )
  else()
    set( CUDA_build_configuration "${CMAKE_BUILD_TYPE}")
  endif()

  # Initialize our list of includes with the user ones followed by the CUDA system ones.
  set(CUDA_NVCC_INCLUDE_ARGS ${CUDA_NVCC_INCLUDE_ARGS_USER} "-I${CUDA_INCLUDE_DIRS}")
  # Get the include directories for this directory and use them for our nvcc command.
  # Remove duplicate entries which may be present since include_directories
  # in CMake >= 2.8.8 does not remove them.
  get_directory_property(CUDA_NVCC_INCLUDE_DIRECTORIES INCLUDE_DIRECTORIES)
  list(REMOVE_DUPLICATES CUDA_NVCC_INCLUDE_DIRECTORIES)
  if(CUDA_NVCC_INCLUDE_DIRECTORIES)
    foreach(dir ${CUDA_NVCC_INCLUDE_DIRECTORIES})
      list(APPEND CUDA_NVCC_INCLUDE_ARGS -I${dir})
    endforeach()
  endif()

  # Reset these variables
  set(CUDA_WRAP_OPTION_NVCC_FLAGS)
  foreach(config ${CUDA_configuration_types})
    string(TOUPPER ${config} config_upper)
    set(CUDA_WRAP_OPTION_NVCC_FLAGS_${config_upper})
  endforeach()

  CUDA_GET_SOURCES_AND_OPTIONS(_cuda_wrap_sources _cuda_wrap_cmake_options _cuda_wrap_options ${ARGN})
  CUDA_PARSE_NVCC_OPTIONS(CUDA_WRAP_OPTION_NVCC_FLAGS ${_cuda_wrap_options})

  # Figure out if we are building a shared library.  BUILD_SHARED_LIBS is
  # respected in CUDA_ADD_LIBRARY.
  set(_cuda_build_shared_libs FALSE)
  # SHARED, MODULE
  list(FIND _cuda_wrap_cmake_options SHARED _cuda_found_SHARED)
  list(FIND _cuda_wrap_cmake_options MODULE _cuda_found_MODULE)
  if(_cuda_found_SHARED GREATER -1 OR _cuda_found_MODULE GREATER -1)
    set(_cuda_build_shared_libs TRUE)
  endif()
  # STATIC
  list(FIND _cuda_wrap_cmake_options STATIC _cuda_found_STATIC)
  if(_cuda_found_STATIC GREATER -1)
    set(_cuda_build_shared_libs FALSE)
  endif()

  # CUDA_HOST_FLAGS
  if(_cuda_build_shared_libs)
    # If we are setting up code for a shared library, then we need to add extra flags for
    # compiling objects for shared libraries.
    set(CUDA_HOST_SHARED_FLAGS ${CMAKE_SHARED_LIBRARY_${CUDA_C_OR_CXX}_FLAGS})
  else()
    set(CUDA_HOST_SHARED_FLAGS)
  endif()
  # Only add the CMAKE_{C,CXX}_FLAGS if we are propagating host flags.  We
  # always need to set the SHARED_FLAGS, though.
  if(CUDA_PROPAGATE_HOST_FLAGS)
    set(_cuda_host_flags "set(CMAKE_HOST_FLAGS ${CMAKE_${CUDA_C_OR_CXX}_FLAGS} ${CUDA_HOST_SHARED_FLAGS})")
  else()
    set(_cuda_host_flags "set(CMAKE_HOST_FLAGS ${CUDA_HOST_SHARED_FLAGS})")
  endif()

  set(_cuda_nvcc_flags_config "# Build specific configuration flags")
  # Loop over all the configuration types to generate appropriate flags for run_nvcc.cmake
  foreach(config ${CUDA_configuration_types})
    string(TOUPPER ${config} config_upper)
    # CMAKE_FLAGS are strings and not lists.  By not putting quotes around CMAKE_FLAGS
    # we convert the strings to lists (like we want).

    if(CUDA_PROPAGATE_HOST_FLAGS)
      # nvcc chokes on -g3 in versions previous to 3.0, so replace it with -g
      set(_cuda_fix_g3 FALSE)

      if(CMAKE_COMPILER_IS_GNUCC)
        if (CUDA_VERSION VERSION_LESS  "3.0" OR
            CUDA_VERSION VERSION_EQUAL "4.1" OR
            CUDA_VERSION VERSION_EQUAL "4.2"
            )
          set(_cuda_fix_g3 TRUE)
        endif()
      endif()
      if(_cuda_fix_g3)
        string(REPLACE "-g3" "-g" _cuda_C_FLAGS "${CMAKE_${CUDA_C_OR_CXX}_FLAGS_${config_upper}}")
      else()
        set(_cuda_C_FLAGS "${CMAKE_${CUDA_C_OR_CXX}_FLAGS_${config_upper}}")
      endif()

      set(_cuda_host_flags "${_cuda_host_flags}\nset(CMAKE_HOST_FLAGS_${config_upper} ${_cuda_C_FLAGS})")
    endif()

    # Note that if we ever want CUDA_NVCC_FLAGS_<CONFIG> to be string (instead of a list
    # like it is currently), we can remove the quotes around the
    # ${CUDA_NVCC_FLAGS_${config_upper}} variable like the CMAKE_HOST_FLAGS_<CONFIG> variable.
    set(_cuda_nvcc_flags_config "${_cuda_nvcc_flags_config}\nset(CUDA_NVCC_FLAGS_${config_upper} ${CUDA_NVCC_FLAGS_${config_upper}} ;; ${CUDA_WRAP_OPTION_NVCC_FLAGS_${config_upper}})")
  endforeach()

  # Process the C++11 flag.  If the host sets the flag, we need to add it to nvcc and
  # remove it from the host. This is because -Xcompile -std=c++ will choke nvcc (it uses
  # the C preprocessor).  In order to get this to work correctly, we need to use nvcc's
  # specific c++11 flag.
  if( "${_cuda_host_flags}" MATCHES "-std=c\\+\\+11")
    # Add the c++11 flag to nvcc if it isn't already present.  Note that we only look at
    # the main flag instead of the configuration specific flags.
    if( NOT "${CUDA_NVCC_FLAGS}" MATCHES "-std;c\\+\\+11" )
      list(APPEND nvcc_flags --std c++11)
    endif()
    string(REGEX REPLACE "[-]+std=c\\+\\+11" "" _cuda_host_flags "${_cuda_host_flags}")
  endif()

  # Get the list of definitions from the directory property
  get_directory_property(CUDA_NVCC_DEFINITIONS COMPILE_DEFINITIONS)
  if(CUDA_NVCC_DEFINITIONS)
    foreach(_definition ${CUDA_NVCC_DEFINITIONS})
      list(APPEND nvcc_flags "-D${_definition}")
    endforeach()
  endif()

  if(_cuda_build_shared_libs)
    list(APPEND nvcc_flags "-D${cuda_target}_EXPORTS")
  endif()

  # Reset the output variable
  set(_cuda_wrap_generated_files "")

  # Iterate over the macro arguments and create custom
  # commands for all the .cu files.
  foreach(file ${ARGN})
    # Ignore any file marked as a HEADER_FILE_ONLY
    get_source_file_property(_is_header ${file} HEADER_FILE_ONLY)
    # Allow per source file overrides of the format.  Also allows compiling non-.cu files.
    get_source_file_property(_cuda_source_format ${file} CUDA_SOURCE_PROPERTY_FORMAT)
    if((${file} MATCHES "\\.cu$" OR _cuda_source_format) AND NOT _is_header)

      if(NOT _cuda_source_format)
        set(_cuda_source_format ${format})
      endif()
      # If file isn't a .cu file, we need to tell nvcc to treat it as such.
      if(NOT ${file} MATCHES "\\.cu$")
        set(cuda_language_flag -x=cu)
      else()
        set(cuda_language_flag)
      endif()

      if( ${_cuda_source_format} MATCHES "OBJ")
        set( cuda_compile_to_external_module OFF )
      else()
        set( cuda_compile_to_external_module ON )
        if( ${_cuda_source_format} MATCHES "PTX" )
          set( cuda_compile_to_external_module_type "ptx" )
        elseif( ${_cuda_source_format} MATCHES "CUBIN")
          set( cuda_compile_to_external_module_type "cubin" )
        elseif( ${_cuda_source_format} MATCHES "FATBIN")
          set( cuda_compile_to_external_module_type "fatbin" )
        else()
          message( FATAL_ERROR "Invalid format flag passed to CUDA_WRAP_SRCS or set with CUDA_SOURCE_PROPERTY_FORMAT file property for file '${file}': '${_cuda_source_format}'.  Use OBJ, PTX, CUBIN or FATBIN.")
        endif()
      endif()

      if(cuda_compile_to_external_module)
        # Don't use any of the host compilation flags for PTX targets.
        set(CUDA_HOST_FLAGS)
        set(CUDA_NVCC_FLAGS_CONFIG)
      else()
        set(CUDA_HOST_FLAGS ${_cuda_host_flags})
        set(CUDA_NVCC_FLAGS_CONFIG ${_cuda_nvcc_flags_config})
      endif()

      # Determine output directory
      cuda_compute_build_path("${file}" cuda_build_path)
      set(cuda_compile_intermediate_directory "${CMAKE_CURRENT_BINARY_DIR}/CMakeFiles/${cuda_target}.dir/${cuda_build_path}")
      if(CUDA_GENERATED_OUTPUT_DIR)
        set(cuda_compile_output_dir "${CUDA_GENERATED_OUTPUT_DIR}")
      else()
        if ( cuda_compile_to_external_module )
          set(cuda_compile_output_dir "${CMAKE_CURRENT_BINARY_DIR}")
        else()
          set(cuda_compile_output_dir "${cuda_compile_intermediate_directory}")
        endif()
      endif()

      # Add a custom target to generate a c or ptx file. ######################

      get_filename_component( basename ${file} NAME )
      if( cuda_compile_to_external_module )
        set(generated_file_path "${cuda_compile_output_dir}")
        set(generated_file_basename "${cuda_target}_generated_${basename}.${cuda_compile_to_external_module_type}")
        set(format_flag "-${cuda_compile_to_external_module_type}")
        file(MAKE_DIRECTORY "${cuda_compile_output_dir}")
      else()
        set(generated_file_path "${cuda_compile_output_dir}/${CMAKE_CFG_INTDIR}")
        set(generated_file_basename "${cuda_target}_generated_${basename}${generated_extension}")
        if(CUDA_SEPARABLE_COMPILATION)
          set(format_flag "-dc")
        else()
          set(format_flag "-c")
        endif()
      endif()

      # Set all of our file names.  Make sure that whatever filenames that have
      # generated_file_path in them get passed in through as a command line
      # argument, so that the ${CMAKE_CFG_INTDIR} gets expanded at run time
      # instead of configure time.
      set(generated_file "${generated_file_path}/${generated_file_basename}")
      set(cmake_dependency_file "${cuda_compile_intermediate_directory}/${generated_file_basename}.depend")
      set(NVCC_generated_dependency_file "${cuda_compile_intermediate_directory}/${generated_file_basename}.NVCC-depend")
      set(generated_cubin_file "${generated_file_path}/${generated_file_basename}.cubin.txt")
      set(custom_target_script "${cuda_compile_intermediate_directory}/${generated_file_basename}.cmake")

      # Setup properties for obj files:
      if( NOT cuda_compile_to_external_module )
        set_source_files_properties("${generated_file}"
          PROPERTIES
          EXTERNAL_OBJECT true # This is an object file not to be compiled, but only be linked.
          )
      endif()

      # Don't add CMAKE_CURRENT_SOURCE_DIR if the path is already an absolute path.
      get_filename_component(file_path "${file}" PATH)
      if(IS_ABSOLUTE "${file_path}")
        set(source_file "${file}")
      else()
        set(source_file "${CMAKE_CURRENT_SOURCE_DIR}/${file}")
      endif()

      if( NOT cuda_compile_to_external_module AND CUDA_SEPARABLE_COMPILATION)
        list(APPEND ${cuda_target}_SEPARABLE_COMPILATION_OBJECTS "${generated_file}")
      endif()

      # Bring in the dependencies.  Creates a variable CUDA_NVCC_DEPEND #######
      cuda_include_nvcc_dependencies(${cmake_dependency_file})

      # Convience string for output ###########################################
      if(CUDA_BUILD_EMULATION)
        set(cuda_build_type "Emulation")
      else()
        set(cuda_build_type "Device")
      endif()

      # Build the NVCC made dependency file ###################################
      set(build_cubin OFF)
      if ( NOT CUDA_BUILD_EMULATION AND CUDA_BUILD_CUBIN )
         if ( NOT cuda_compile_to_external_module )
           set ( build_cubin ON )
         endif()
      endif()

      # Configure the build script
      configure_file("${CUDA_run_nvcc}" "${custom_target_script}" @ONLY)


      # So if a user specifies the same cuda file as input more than once, you
      # can have bad things happen with dependencies.  Here we check an option
      # to see if this is the behavior they want.
      if(CUDA_ATTACH_VS_BUILD_RULE_TO_CUDA_FILE)
        set(main_dep MAIN_DEPENDENCY ${source_file})
      else()
        set(main_dep DEPENDS ${source_file})
      endif()

      if(CUDA_VERBOSE_BUILD)
        set(verbose_output ON)
      elseif(CMAKE_GENERATOR MATCHES "Makefiles")
        set(verbose_output "$(VERBOSE)")
      else()
        set(verbose_output OFF)
      endif()

      # Create up the comment string
      file(RELATIVE_PATH generated_file_relative_path "${CMAKE_BINARY_DIR}" "${generated_file}")
      if(cuda_compile_to_external_module)
        set(cuda_build_comment_string "Building NVCC ${cuda_compile_to_external_module_type} file ${generated_file_relative_path}")
      else()
        set(cuda_build_comment_string "Building NVCC (${cuda_build_type}) object ${generated_file_relative_path}")
      endif()

      set(_verbatim VERBATIM)
      if(ccbin_flags MATCHES "\\$\\(VCInstallDir\\)")
        set(_verbatim "")
      endif()

      # Build the generated file and dependency file ##########################
      add_custom_command(
        OUTPUT ${generated_file}
        # These output files depend on the source_file and the contents of cmake_dependency_file
        ${main_dep}
        DEPENDS ${CUDA_NVCC_DEPEND}
        DEPENDS ${custom_target_script}
        # Make sure the output directory exists before trying to write to it.
        COMMAND ${CMAKE_COMMAND} -E make_directory "${generated_file_path}"
        COMMAND ${CMAKE_COMMAND} ARGS
          -D verbose:BOOL=${verbose_output}
          ${ccbin_flags}
          -D build_configuration:STRING=${CUDA_build_configuration}
          -D "generated_file:STRING=${generated_file}"
          -D "generated_cubin_file:STRING=${generated_cubin_file}"
          -P "${custom_target_script}"
        WORKING_DIRECTORY "${cuda_compile_intermediate_directory}"
        COMMENT "${cuda_build_comment_string}"
        ${_verbatim}
        )

      # Make sure the build system knows the file is generated.
      set_source_files_properties(${generated_file} PROPERTIES GENERATED TRUE)

      list(APPEND _cuda_wrap_generated_files ${generated_file})

      # Add the other files that we want cmake to clean on a cleanup ##########
      list(APPEND CUDA_ADDITIONAL_CLEAN_FILES "${cmake_dependency_file}")
      list(REMOVE_DUPLICATES CUDA_ADDITIONAL_CLEAN_FILES)
      set(CUDA_ADDITIONAL_CLEAN_FILES ${CUDA_ADDITIONAL_CLEAN_FILES} CACHE INTERNAL "List of intermediate files that are part of the cuda dependency scanning.")

    endif()
  endforeach()

  # Set the return parameter
  set(${generated_files} ${_cuda_wrap_generated_files})
endmacro()

function(_cuda_get_important_host_flags important_flags flag_string)
  if(CMAKE_GENERATOR MATCHES "Visual Studio")
    string(REGEX MATCHALL "/M[DT][d]?" flags "${flag_string}")
    list(APPEND ${important_flags} ${flags})
  else()
    string(REGEX MATCHALL "-fPIC" flags "${flag_string}")
    list(APPEND ${important_flags} ${flags})
  endif()
  set(${important_flags} ${${important_flags}} PARENT_SCOPE)
endfunction()

###############################################################################
###############################################################################
# Separable Compilation Link
###############################################################################
###############################################################################

# Compute the filename to be used by CUDA_LINK_SEPARABLE_COMPILATION_OBJECTS
function(CUDA_COMPUTE_SEPARABLE_COMPILATION_OBJECT_FILE_NAME output_file_var cuda_target object_files)
  if (object_files)
    set(generated_extension ${CMAKE_${CUDA_C_OR_CXX}_OUTPUT_EXTENSION})
    set(output_file "${CMAKE_CURRENT_BINARY_DIR}/CMakeFiles/${cuda_target}.dir/${CMAKE_CFG_INTDIR}/${cuda_target}_intermediate_link${generated_extension}")
  else()
    set(output_file)
  endif()

  set(${output_file_var} "${output_file}" PARENT_SCOPE)
endfunction()

# Setup the build rule for the separable compilation intermediate link file.
function(CUDA_LINK_SEPARABLE_COMPILATION_OBJECTS output_file cuda_target options object_files)
  if (object_files)

    set_source_files_properties("${output_file}"
      PROPERTIES
      EXTERNAL_OBJECT TRUE # This is an object file not to be compiled, but only
                           # be linked.
      GENERATED TRUE       # This file is generated during the build
      )

    # For now we are ignoring all the configuration specific flags.
    set(nvcc_flags)
    CUDA_PARSE_NVCC_OPTIONS(nvcc_flags ${options})
    if(CUDA_64_BIT_DEVICE_CODE)
      list(APPEND nvcc_flags -m64)
    else()
      list(APPEND nvcc_flags -m32)
    endif()
    # If -ccbin, --compiler-bindir has been specified, don't do anything.  Otherwise add it here.
    list( FIND nvcc_flags "-ccbin" ccbin_found0 )
    list( FIND nvcc_flags "--compiler-bindir" ccbin_found1 )
    if( ccbin_found0 LESS 0 AND ccbin_found1 LESS 0 AND CUDA_HOST_COMPILER )
      # Match VERBATIM check below.
      if(CUDA_HOST_COMPILER MATCHES "\\$\\(VCInstallDir\\)")
        list(APPEND nvcc_flags -ccbin "\"${CUDA_HOST_COMPILER}\"")
      else()
        list(APPEND nvcc_flags -ccbin "${CUDA_HOST_COMPILER}")
      endif()
    endif()

    # Create a list of flags specified by CUDA_NVCC_FLAGS_${CONFIG} and CMAKE_${CUDA_C_OR_CXX}_FLAGS*
    set(config_specific_flags)
    set(flags)
    foreach(config ${CUDA_configuration_types})
      string(TOUPPER ${config} config_upper)
      # Add config specific flags
      foreach(f ${CUDA_NVCC_FLAGS_${config_upper}})
        list(APPEND config_specific_flags $<$<CONFIG:${config}>:${f}>)
      endforeach()
      set(important_host_flags)
      _cuda_get_important_host_flags(important_host_flags "${CMAKE_${CUDA_C_OR_CXX}_FLAGS_${config_upper}}")
      foreach(f ${important_host_flags})
        list(APPEND flags $<$<CONFIG:${config}>:-Xcompiler> $<$<CONFIG:${config}>:${f}>)
      endforeach()
    endforeach()
    # Add CMAKE_${CUDA_C_OR_CXX}_FLAGS
    set(important_host_flags)
    _cuda_get_important_host_flags(important_host_flags "${CMAKE_${CUDA_C_OR_CXX}_FLAGS}")
    foreach(f ${important_host_flags})
      list(APPEND flags -Xcompiler ${f})
    endforeach()

    # Add our general CUDA_NVCC_FLAGS with the configuration specifig flags
    set(nvcc_flags ${CUDA_NVCC_FLAGS} ${config_specific_flags} ${nvcc_flags})

    file(RELATIVE_PATH output_file_relative_path "${CMAKE_BINARY_DIR}" "${output_file}")

    # Some generators don't handle the multiple levels of custom command
    # dependencies correctly (obj1 depends on file1, obj2 depends on obj1), so
    # we work around that issue by compiling the intermediate link object as a
    # pre-link custom command in that situation.
    set(do_obj_build_rule TRUE)
    if (MSVC_VERSION GREATER 1599 AND MSVC_VERSION LESS 1800)
      # VS 2010 and 2012 have this problem.
      set(do_obj_build_rule FALSE)
    endif()

    set(_verbatim VERBATIM)
    if(nvcc_flags MATCHES "\\$\\(VCInstallDir\\)")
      set(_verbatim "")
    endif()

    if (do_obj_build_rule)
      add_custom_command(
        OUTPUT ${output_file}
        DEPENDS ${object_files}
        COMMAND ${CUDA_NVCC_EXECUTABLE} ${nvcc_flags} -dlink ${object_files} -o ${output_file}
        ${flags}
        COMMENT "Building NVCC intermediate link file ${output_file_relative_path}"
        ${_verbatim}
        )
    else()
      get_filename_component(output_file_dir "${output_file}" DIRECTORY)
      add_custom_command(
        TARGET ${cuda_target}
        PRE_LINK
        COMMAND ${CMAKE_COMMAND} -E echo "Building NVCC intermediate link file ${output_file_relative_path}"
        COMMAND ${CMAKE_COMMAND} -E make_directory "${output_file_dir}"
        COMMAND ${CUDA_NVCC_EXECUTABLE} ${nvcc_flags} ${flags} -dlink ${object_files} -o "${output_file}"
        ${_verbatim}
        )
    endif()
 endif()
endfunction()

###############################################################################
###############################################################################
# ADD LIBRARY
###############################################################################
###############################################################################
macro(CUDA_ADD_LIBRARY cuda_target)

  CUDA_ADD_CUDA_INCLUDE_ONCE()

  # Separate the sources from the options
  CUDA_GET_SOURCES_AND_OPTIONS(_sources _cmake_options _options ${ARGN})
  CUDA_BUILD_SHARED_LIBRARY(_cuda_shared_flag ${ARGN})
  # Create custom commands and targets for each file.
  CUDA_WRAP_SRCS( ${cuda_target} OBJ _generated_files ${_sources}
    ${_cmake_options} ${_cuda_shared_flag}
    OPTIONS ${_options} )

  # Compute the file name of the intermedate link file used for separable
  # compilation.
  CUDA_COMPUTE_SEPARABLE_COMPILATION_OBJECT_FILE_NAME(link_file ${cuda_target} "${${cuda_target}_SEPARABLE_COMPILATION_OBJECTS}")

  # Add the library.
  add_library(${cuda_target} ${_cmake_options}
    ${_generated_files}
    ${_sources}
    ${link_file}
    )

  # Add a link phase for the separable compilation if it has been enabled.  If
  # it has been enabled then the ${cuda_target}_SEPARABLE_COMPILATION_OBJECTS
  # variable will have been defined.
  CUDA_LINK_SEPARABLE_COMPILATION_OBJECTS("${link_file}" ${cuda_target} "${_options}" "${${cuda_target}_SEPARABLE_COMPILATION_OBJECTS}")

  target_link_libraries(${cuda_target}
    ${CUDA_LIBRARIES}
    )

  # We need to set the linker language based on what the expected generated file
  # would be. CUDA_C_OR_CXX is computed based on CUDA_HOST_COMPILATION_CPP.
  set_target_properties(${cuda_target}
    PROPERTIES
    LINKER_LANGUAGE ${CUDA_C_OR_CXX}
    )

endmacro()


###############################################################################
###############################################################################
# ADD EXECUTABLE
###############################################################################
###############################################################################
macro(CUDA_ADD_EXECUTABLE cuda_target)

  CUDA_ADD_CUDA_INCLUDE_ONCE()

  # Separate the sources from the options
  CUDA_GET_SOURCES_AND_OPTIONS(_sources _cmake_options _options ${ARGN})
  # Create custom commands and targets for each file.
  CUDA_WRAP_SRCS( ${cuda_target} OBJ _generated_files ${_sources} OPTIONS ${_options} )

  # Compute the file name of the intermedate link file used for separable
  # compilation.
  CUDA_COMPUTE_SEPARABLE_COMPILATION_OBJECT_FILE_NAME(link_file ${cuda_target} "${${cuda_target}_SEPARABLE_COMPILATION_OBJECTS}")

  # Add the library.
  add_executable(${cuda_target} ${_cmake_options}
    ${_generated_files}
    ${_sources}
    ${link_file}
    )

  # Add a link phase for the separable compilation if it has been enabled.  If
  # it has been enabled then the ${cuda_target}_SEPARABLE_COMPILATION_OBJECTS
  # variable will have been defined.
  CUDA_LINK_SEPARABLE_COMPILATION_OBJECTS("${link_file}" ${cuda_target} "${_options}" "${${cuda_target}_SEPARABLE_COMPILATION_OBJECTS}")

  target_link_libraries(${cuda_target}
    ${CUDA_LIBRARIES}
    )

  # We need to set the linker language based on what the expected generated file
  # would be. CUDA_C_OR_CXX is computed based on CUDA_HOST_COMPILATION_CPP.
  set_target_properties(${cuda_target}
    PROPERTIES
    LINKER_LANGUAGE ${CUDA_C_OR_CXX}
    )

endmacro()


###############################################################################
###############################################################################
# (Internal) helper for manually added cuda source files with specific targets
###############################################################################
###############################################################################
macro(cuda_compile_base cuda_target format generated_files)

  # Separate the sources from the options
  CUDA_GET_SOURCES_AND_OPTIONS(_sources _cmake_options _options ${ARGN})
  # Create custom commands and targets for each file.
  CUDA_WRAP_SRCS( ${cuda_target} ${format} _generated_files ${_sources} ${_cmake_options}
    OPTIONS ${_options} )

  set( ${generated_files} ${_generated_files})

endmacro()

###############################################################################
###############################################################################
# CUDA COMPILE
###############################################################################
###############################################################################
macro(CUDA_COMPILE generated_files)
  cuda_compile_base(cuda_compile OBJ ${generated_files} ${ARGN})
endmacro()

###############################################################################
###############################################################################
# CUDA COMPILE PTX
###############################################################################
###############################################################################
macro(CUDA_COMPILE_PTX generated_files)
  cuda_compile_base(cuda_compile_ptx PTX ${generated_files} ${ARGN})
endmacro()

###############################################################################
###############################################################################
# CUDA COMPILE FATBIN
###############################################################################
###############################################################################
macro(CUDA_COMPILE_FATBIN generated_files)
  cuda_compile_base(cuda_compile_fatbin FATBIN ${generated_files} ${ARGN})
endmacro()

###############################################################################
###############################################################################
# CUDA COMPILE CUBIN
###############################################################################
###############################################################################
macro(CUDA_COMPILE_CUBIN generated_files)
  cuda_compile_base(cuda_compile_cubin CUBIN ${generated_files} ${ARGN})
endmacro()


###############################################################################
###############################################################################
# CUDA ADD CUFFT TO TARGET
###############################################################################
###############################################################################
macro(CUDA_ADD_CUFFT_TO_TARGET target)
  if (CUDA_BUILD_EMULATION)
    target_link_libraries(${target} ${CUDA_cufftemu_LIBRARY})
  else()
    target_link_libraries(${target} ${CUDA_cufft_LIBRARY})
  endif()
endmacro()

###############################################################################
###############################################################################
# CUDA ADD CUBLAS TO TARGET
###############################################################################
###############################################################################
macro(CUDA_ADD_CUBLAS_TO_TARGET target)
  if (CUDA_BUILD_EMULATION)
    target_link_libraries(${target} ${CUDA_cublasemu_LIBRARY})
  else()
    target_link_libraries(${target} ${CUDA_cublas_LIBRARY} ${CUDA_cublas_device_LIBRARY})
  endif()
endmacro()

###############################################################################
###############################################################################
# CUDA BUILD CLEAN TARGET
###############################################################################
###############################################################################
macro(CUDA_BUILD_CLEAN_TARGET)
  # Call this after you add all your CUDA targets, and you will get a convience
  # target.  You should also make clean after running this target to get the
  # build system to generate all the code again.

  set(cuda_clean_target_name clean_cuda_depends)
  if (CMAKE_GENERATOR MATCHES "Visual Studio")
    string(TOUPPER ${cuda_clean_target_name} cuda_clean_target_name)
  endif()
  add_custom_target(${cuda_clean_target_name}
    COMMAND ${CMAKE_COMMAND} -E remove ${CUDA_ADDITIONAL_CLEAN_FILES})

  # Clear out the variable, so the next time we configure it will be empty.
  # This is useful so that the files won't persist in the list after targets
  # have been removed.
  set(CUDA_ADDITIONAL_CLEAN_FILES "" CACHE INTERNAL "List of intermediate files that are part of the cuda dependency scanning.")
endmacro()



================================================
FILE: cmake/3.6/Modules/FindCUDA/make2cmake.cmake
================================================
#  James Bigler, NVIDIA Corp (nvidia.com - jbigler)
#  Abe Stephens, SCI Institute -- http://www.sci.utah.edu/~abe/FindCuda.html
#
#  Copyright (c) 2008 - 2009 NVIDIA Corporation.  All rights reserved.
#
#  Copyright (c) 2007-2009
#  Scientific Computing and Imaging Institute, University of Utah
#
#  This code is licensed under the MIT License.  See the FindCUDA.cmake script
#  for the text of the license.

# The MIT License
#
# License for the specific language governing rights and limitations under
# Permission is hereby granted, free of charge, to any person obtaining a
# copy of this software and associated documentation files (the "Software"),
# to deal in the Software without restriction, including without limitation
# the rights to use, copy, modify, merge, publish, distribute, sublicense,
# and/or sell copies of the Software, and to permit persons to whom the
# Software is furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included
# in all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
# DEALINGS IN THE SOFTWARE.
#

#######################################################################
# This converts a file written in makefile syntax into one that can be included
# by CMake.

# Input variables
#
# verbose:BOOL=<>          OFF: Be as quiet as possible (default)
#                          ON : Extra output
#
# input_file:FILEPATH=<>   Path to dependecy file in makefile format
#
# output_file:FILEPATH=<>  Path to file with dependencies in CMake readable variable
#

file(READ ${input_file} depend_text)

if (NOT "${depend_text}" STREQUAL "")

  # message("FOUND DEPENDS")

  string(REPLACE "\\ " " " depend_text ${depend_text})

  # This works for the nvcc -M generated dependency files.
  string(REGEX REPLACE "^.* : " "" depend_text ${depend_text})
  string(REGEX REPLACE "[ \\\\]*\n" ";" depend_text ${depend_text})

  set(dependency_list "")

  foreach(file ${depend_text})

    string(REGEX REPLACE "^ +" "" file ${file})

    # OK, now if we had a UNC path, nvcc has a tendency to only output the first '/'
    # instead of '//'.  Here we will test to see if the file exists, if it doesn't then
    # try to prepend another '/' to the path and test again.  If it still fails remove the
    # path.

    if(NOT EXISTS "${file}")
      if (EXISTS "/${file}")
        set(file "/${file}")
      else()
        if(verbose)
          message(WARNING " Removing non-existent dependency file: ${file}")
        endif()
        set(file "")
      endif()
    endif()

    # Make sure we check to see if we have a file, before asking if it is not a directory.
    # if(NOT IS_DIRECTORY "") will return TRUE.
    if(file AND NOT IS_DIRECTORY "${file}")
      # If softlinks start to matter, we should change this to REALPATH.  For now we need
      # to flatten paths, because nvcc can generate stuff like /bin/../include instead of
      # just /include.
      get_filename_component(file_absolute "${file}" ABSOLUTE)
      list(APPEND dependency_list "${file_absolute}")
    endif()

  endforeach()

else()
  # message("FOUND NO DEPENDS")
endif()

# Remove the duplicate entries and sort them.
list(REMOVE_DUPLICATES dependency_list)
list(SORT dependency_list)

foreach(file ${dependency_list})
  set(cuda_nvcc_depend "${cuda_nvcc_depend} \"${file}\"\n")
endforeach()

file(WRITE ${output_file} "# Generated by: make2cmake.cmake\nSET(CUDA_NVCC_DEPEND\n ${cuda_nvcc_depend})\n\n")



================================================
FILE: cmake/3.6/Modules/FindCUDA/parse_cubin.cmake
================================================
#  James Bigler, NVIDIA Corp (nvidia.com - jbigler)
#  Abe Stephens, SCI Institute -- http://www.sci.utah.edu/~abe/FindCuda.html
#
#  Copyright (c) 2008 - 2009 NVIDIA Corporation.  All rights reserved.
#
#  Copyright (c) 2007-2009
#  Scientific Computing and Imaging Institute, University of Utah
#
#  This code is licensed under the MIT License.  See the FindCUDA.cmake script
#  for the text of the license.

# The MIT License
#
# License for the specific language governing rights and limitations under
# Permission is hereby granted, free of charge, to any person obtaining a
# copy of this software and associated documentation files (the "Software"),
# to deal in the Software without restriction, including without limitation
# the rights to use, copy, modify, merge, publish, distribute, sublicense,
# and/or sell copies of the Software, and to permit persons to whom the
# Software is furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included
# in all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
# DEALINGS IN THE SOFTWARE.
#

#######################################################################
# Parses a .cubin file produced by nvcc and reports statistics about the file.


file(READ ${input_file} file_text)

if (NOT "${file_text}" STREQUAL "")

  string(REPLACE ";" "\\;" file_text ${file_text})
  string(REPLACE "\ncode" ";code" file_text ${file_text})

  list(LENGTH file_text len)

  foreach(line ${file_text})

    # Only look at "code { }" blocks.
    if(line MATCHES "^code")

      # Break into individual lines.
      string(REGEX REPLACE "\n" ";" line ${line})

      foreach(entry ${line})

        # Extract kernel names.
        if (${entry} MATCHES "[^g]name = ([^ ]+)")
          set(entry "${CMAKE_MATCH_1}")

          # Check to see if the kernel name starts with "_"
          set(skip FALSE)
          # if (${entry} MATCHES "^_")
            # Skip the rest of this block.
            # message("Skipping ${entry}")
            # set(skip TRUE)
          # else ()
            message("Kernel:    ${entry}")
          # endif ()

        endif()

        # Skip the rest of the block if necessary
        if(NOT skip)

          # Registers
          if (${entry} MATCHES "reg([ ]+)=([ ]+)([^ ]+)")
            set(entry "${CMAKE_MATCH_3}")
            message("Registers: ${entry}")
          endif()

          # Local memory
          if (${entry} MATCHES "lmem([ ]+)=([ ]+)([^ ]+)")
            set(entry "${CMAKE_MATCH_3}")
            message("Local:     ${entry}")
          endif()

          # Shared memory
          if (${entry} MATCHES "smem([ ]+)=([ ]+)([^ ]+)")
            set(entry "${CMAKE_MATCH_3}")
            message("Shared:    ${entry}")
          endif()

          if (${entry} MATCHES "^}")
            message("")
          endif()

        endif()


      endforeach()

    endif()

  endforeach()

else()
  # message("FOUND NO DEPENDS")
endif()





================================================
FILE: cmake/3.6/Modules/FindCUDA/run_nvcc.cmake
================================================
#  James Bigler, NVIDIA Corp (nvidia.com - jbigler)
#
#  Copyright (c) 2008 - 2009 NVIDIA Corporation.  All rights reserved.
#
#  This code is licensed under the MIT License.  See the FindCUDA.cmake script
#  for the text of the license.

# The MIT License
#
# License for the specific language governing rights and limitations under
# Permission is hereby granted, free of charge, to any person obtaining a
# copy of this software and associated documentation files (the "Software"),
# to deal in the Software without restriction, including without limitation
# the rights to use, copy, modify, merge, publish, distribute, sublicense,
# and/or sell copies of the Software, and to permit persons to whom the
# Software is furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included
# in all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
# DEALINGS IN THE SOFTWARE.


##########################################################################
# This file runs the nvcc commands to produce the desired output file along with
# the dependency file needed by CMake to compute dependencies.  In addition the
# file checks the output of each command and if the command fails it deletes the
# output files.

# Input variables
#
# verbose:BOOL=<>          OFF: Be as quiet as possible (default)
#                          ON : Describe each step
#
# build_configuration:STRING=<> Typically one of Debug, MinSizeRel, Release, or
#                               RelWithDebInfo, but it should match one of the
#                               entries in CUDA_HOST_FLAGS. This is the build
#                               configuration used when compiling the code.  If
#                               blank or unspecified Debug is assumed as this is
#                               what CMake does.
#
# generated_file:STRING=<> File to generate.  This argument must be passed in.
#
# generated_cubin_file:STRING=<> File to generate.  This argument must be passed
#                                                   in if build_cubin is true.

if(NOT generated_file)
  message(FATAL_ERROR "You must specify generated_file on the command line")
endif()

# Set these up as variables to make reading the generated file easier
set(CMAKE_COMMAND "@CMAKE_COMMAND@") # path
set(source_file "@source_file@") # path
set(NVCC_generated_dependency_file "@NVCC_generated_dependency_file@") # path
set(cmake_dependency_file "@cmake_dependency_file@") # path
set(CUDA_make2cmake "@CUDA_make2cmake@") # path
set(CUDA_parse_cubin "@CUDA_parse_cubin@") # path
set(build_cubin @build_cubin@) # bool
set(CUDA_HOST_COMPILER "@CUDA_HOST_COMPILER@") # path
# We won't actually use these variables for now, but we need to set this, in
# order to force this file to be run again if it changes.
set(generated_file_path "@generated_file_path@") # path
set(generated_file_internal "@generated_file@") # path
set(generated_cubin_file_internal "@generated_cubin_file@") # path

set(CUDA_NVCC_EXECUTABLE "@CUDA_NVCC_EXECUTABLE@") # path
set(CUDA_NVCC_FLAGS @CUDA_NVCC_FLAGS@ ;; @CUDA_WRAP_OPTION_NVCC_FLAGS@) # list
@CUDA_NVCC_FLAGS_CONFIG@
set(nvcc_flags @nvcc_flags@) # list
set(CUDA_NVCC_INCLUDE_ARGS "@CUDA_NVCC_INCLUDE_ARGS@") # list (needs to be in quotes to handle spaces properly).
set(format_flag "@format_flag@") # string
set(cuda_language_flag @cuda_language_flag@) # list

if(build_cubin AND NOT generated_cubin_file)
  message(FATAL_ERROR "You must specify generated_cubin_file on the command line")
endif()

# This is the list of host compilation flags.  It C or CXX should already have
# been chosen by FindCUDA.cmake.
@CUDA_HOST_FLAGS@

# Take the compiler flags and package them up to be sent to the compiler via -Xcompiler
set(nvcc_host_compiler_flags "")
# If we weren't given a build_configuration, use Debug.
if(NOT build_configuration)
  set(build_configuration Debug)
endif()
string(TOUPPER "${build_configuration}" build_configuration)
#message("CUDA_NVCC_HOST_COMPILER_FLAGS = ${CUDA_NVCC_HOST_COMPILER_FLAGS}")
foreach(flag ${CMAKE_HOST_FLAGS} ${CMAKE_HOST_FLAGS_${build_configuration}})
  # Extra quotes are added around each flag to help nvcc parse out flags with spaces.
  set(nvcc_host_compiler_flags "${nvcc_host_compiler_flags},\"${flag}\"")
endforeach()
if (nvcc_host_compiler_flags)
  set(nvcc_host_compiler_flags "-Xcompiler" ${nvcc_host_compiler_flags})
endif()
#message("nvcc_host_compiler_flags = \"${nvcc_host_compiler_flags}\"")
# Add the build specific configuration flags
list(APPEND CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS_${build_configuration}})

# Any -ccbin existing in CUDA_NVCC_FLAGS gets highest priority
list( FIND CUDA_NVCC_FLAGS "-ccbin" ccbin_found0 )
list( FIND CUDA_NVCC_FLAGS "--compiler-bindir" ccbin_found1 )
if( ccbin_found0 LESS 0 AND ccbin_found1 LESS 0 AND CUDA_HOST_COMPILER )
  if (CUDA_HOST_COMPILER STREQUAL "$(VCInstallDir)bin" AND DEFINED CCBIN)
    set(CCBIN -ccbin "${CCBIN}")
  else()
    set(CCBIN -ccbin "${CUDA_HOST_COMPILER}")
  endif()
endif()

# cuda_execute_process - Executes a command with optional command echo and status message.
#
#   status  - Status message to print if verbose is true
#   command - COMMAND argument from the usual execute_process argument structure
#   ARGN    - Remaining arguments are the command with arguments
#
#   CUDA_result - return value from running the command
#
# Make this a macro instead of a function, so that things like RESULT_VARIABLE
# and other return variables are present after executing the process.
macro(cuda_execute_process status command)
  set(_command ${command})
  if(NOT "x${_command}" STREQUAL "xCOMMAND")
    message(FATAL_ERROR "Malformed call to cuda_execute_process.  Missing COMMAND as second argument. (command = ${command})")
  endif()
  if(verbose)
    execute_process(COMMAND "${CMAKE_COMMAND}" -E echo -- ${status})
    # Now we need to build up our command string.  We are accounting for quotes
    # and spaces, anything else is left up to the user to fix if they want to
    # copy and paste a runnable command line.
    set(cuda_execute_process_string)
    foreach(arg ${ARGN})
      # If there are quotes, excape them, so they come through.
      string(REPLACE "\"" "\\\"" arg ${arg})
      # Args with spaces need quotes around them to get them to be parsed as a single argument.
      if(arg MATCHES " ")
        list(APPEND cuda_execute_process_string "\"${arg}\"")
      else()
        list(APPEND cuda_execute_process_string ${arg})
      endif()
    endforeach()
    # Echo the command
    execute_process(COMMAND ${CMAKE_COMMAND} -E echo ${cuda_execute_process_string})
  endif()
  # Run the command
  execute_process(COMMAND ${ARGN} RESULT_VARIABLE CUDA_result )
endmacro()

# Delete the target file
cuda_execute_process(
  "Removing ${generated_file}"
  COMMAND "${CMAKE_COMMAND}" -E remove "${generated_file}"
  )

# For CUDA 2.3 and below, -G -M doesn't work, so remove the -G flag
# for dependency generation and hope for the best.
set(depends_CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS}")
set(CUDA_VERSION @CUDA_VERSION@)
if(CUDA_VERSION VERSION_LESS "3.0")
  cmake_policy(PUSH)
  # CMake policy 0007 NEW states that empty list elements are not
  # ignored.  I'm just setting it to avoid the warning that's printed.
  cmake_policy(SET CMP0007 NEW)
  # Note that this will remove all occurances of -G.
  list(REMOVE_ITEM depends_CUDA_NVCC_FLAGS "-G")
  cmake_policy(POP)
endif()

# nvcc doesn't define __CUDACC__ for some reason when generating dependency files.  This
# can cause incorrect dependencies when #including files based on this macro which is
# defined in the generating passes of nvcc invokation.  We will go ahead and manually
# define this for now until a future version fixes this bug.
set(CUDACC_DEFINE -D__CUDACC__)

# Generate the dependency file
cuda_execute_process(
  "Generating dependency file: ${NVCC_generated_dependency_file}"
  COMMAND "${CUDA_NVCC_EXECUTABLE}"
  -M
  ${CUDACC_DEFINE}
  "${source_file}"
  -o "${NVCC_generated_dependency_file}"
  ${CCBIN}
  ${nvcc_flags}
  ${nvcc_host_compiler_flags}
  ${depends_CUDA_NVCC_FLAGS}
  -DNVCC
  ${CUDA_NVCC_INCLUDE_ARGS}
  )

if(CUDA_result)
  message(FATAL_ERROR "Error generating ${generated_file}")
endif()

# Generate the cmake readable dependency file to a temp file.  Don't put the
# quotes just around the filenames for the input_file and output_file variables.
# CMake will pass the quotes through and not be able to find the file.
cuda_execute_process(
  "Generating temporary cmake readable file: ${cmake_dependency_file}.tmp"
  COMMAND "${CMAKE_COMMAND}"
  -D "input_file:FILEPATH=${NVCC_generated_dependency_file}"
  -D "output_file:FILEPATH=${cmake_dependency_file}.tmp"
  -D "verbose=${verbose}"
  -P "${CUDA_make2cmake}"
  )

if(CUDA_result)
  message(FATAL_ERROR "Error generating ${generated_file}")
endif()

# Copy the file if it is different
cuda_execute_process(
  "Copy if different ${cmake_dependency_file}.tmp to ${cmake_dependency_file}"
  COMMAND "${CMAKE_COMMAND}" -E copy_if_different "${cmake_dependency_file}.tmp" "${cmake_dependency_file}"
  )

if(CUDA_result)
  message(FATAL_ERROR "Error generating ${generated_file}")
endif()

# Delete the temporary file
cuda_execute_process(
  "Removing ${cmake_dependency_file}.tmp and ${NVCC_generated_dependency_file}"
  COMMAND "${CMAKE_COMMAND}" -E remove "${cmake_dependency_file}.tmp" "${NVCC_generated_dependency_file}"
  )

if(CUDA_result)
  message(FATAL_ERROR "Error generating ${generated_file}")
endif()

# Generate the code
cuda_execute_process(
  "Generating ${generated_file}"
  COMMAND "${CUDA_NVCC_EXECUTABLE}"
  "${source_file}"
  ${cuda_language_flag}
  ${format_flag} -o "${generated_file}"
  ${CCBIN}
  ${nvcc_flags}
  ${nvcc_host_compiler_flags}
  ${CUDA_NVCC_FLAGS}
  -DNVCC
  ${CUDA_NVCC_INCLUDE_ARGS}
  )

if(CUDA_result)
  # Since nvcc can sometimes leave half done files make sure that we delete the output file.
  cuda_execute_process(
    "Removing ${generated_file}"
    COMMAND "${CMAKE_COMMAND}" -E remove "${generated_file}"
    )
  message(FATAL_ERROR "Error generating file ${generated_file}")
else()
  if(verbose)
    message("Generated ${generated_file} successfully.")
  endif()
endif()

# Cubin resource report commands.
if( build_cubin )
  # Run with -cubin to produce resource usage report.
  cuda_execute_process(
    "Generating ${generated_cubin_file}"
    COMMAND "${CUDA_NVCC_EXECUTABLE}"
    "${source_file}"
    ${CUDA_NVCC_FLAGS}
    ${nvcc_flags}
    ${CCBIN}
    ${nvcc_host_compiler_flags}
    -DNVCC
    -cubin
    -o "${generated_cubin_file}"
    ${CUDA_NVCC_INCLUDE_ARGS}
    )

  # Execute the parser script.
  cuda_execute_process(
    "Executing the parser script"
    COMMAND  "${CMAKE_COMMAND}"
    -D "input_file:STRING=${generated_cubin_file}"
    -P "${CUDA_parse_cubin}"
    )

endif()



================================================
FILE: cmake/3.6/Modules/FindCUDA/select_compute_arch.cmake
================================================
# Synopsis:
#   CUDA_SELECT_NVCC_ARCH_FLAGS(out_variable [target_CUDA_architectures])
#   -- Selects GPU arch flags for nvcc based on target_CUDA_architectures
#      target_CUDA_architectures : Auto | Common | All | LIST(ARCH_AND_PTX ...)
#       - "Auto" detects local machine GPU compute arch at runtime.
#       - "Common" and "All" cover common and entire subsets of architectures
#      ARCH_AND_PTX : NAME | NUM.NUM | NUM.NUM(NUM.NUM) | NUM.NUM+PTX
#      NAME: Fermi Kepler Maxwell Kepler+Tegra Kepler+Tesla Maxwell+Tegra Pascal
#      NUM: Any number. Only those pairs are currently accepted by NVCC though:
#            2.0 2.1 3.0 3.2 3.5 3.7 5.0 5.2 5.3 6.0 6.2
#      Returns LIST of flags to be added to CUDA_NVCC_FLAGS in ${out_variable}
#      Additionally, sets ${out_variable}_readable to the resulting numeric list
#      Example:
#       CUDA_SELECT_NVCC_ARCH_FLAGS(ARCH_FLAGS 3.0 3.5+PTX 5.2(5.0) Maxwell)
#        LIST(APPEND CUDA_NVCC_FLAGS ${ARCH_FLAGS})
#
#      More info on CUDA architectures: https://en.wikipedia.org/wiki/CUDA
#

# This list will be used for CUDA_ARCH_NAME = All option
set(CUDA_KNOWN_GPU_ARCHITECTURES  "Fermi" "Kepler" "Maxwell")

# This list will be used for CUDA_ARCH_NAME = Common option (enabled by default)
set(CUDA_COMMON_GPU_ARCHITECTURES "3.0" "3.5" "5.0")

if (CUDA_VERSION VERSION_GREATER "6.5")
  list(APPEND CUDA_KNOWN_GPU_ARCHITECTURES "Kepler+Tegra" "Kepler+Tesla" "Maxwell+Tegra")
  list(APPEND CUDA_COMMON_GPU_ARCHITECTURES "5.2")
endif ()

if (CUDA_VERSION VERSION_GREATER "7.5")
  list(APPEND CUDA_KNOWN_GPU_ARCHITECTURES "Pascal")
  list(APPEND CUDA_COMMON_GPU_ARCHITECTURES "6.0" "6.1" "6.1+PTX")
else()
  list(APPEND CUDA_COMMON_GPU_ARCHITECTURES "5.2+PTX")
endif ()



################################################################################################
# A function for automatic detection of GPUs installed  (if autodetection is enabled)
# Usage:
#   CUDA_DETECT_INSTALLED_GPUS(OUT_VARIABLE)
#
function(CUDA_DETECT_INSTALLED_GPUS OUT_VARIABLE)
  if(NOT CUDA_GPU_DETECT_OUTPUT)
    set(cufile ${PROJECT_BINARY_DIR}/detect_cuda_archs.cu)

    file(WRITE ${cufile} ""
      "#include <cstdio>\n"
      "int main()\n"
      "{\n"
      "  int count = 0;\n"
      "  if (cudaSuccess != cudaGetDeviceCount(&count)) return -1;\n"
      "  if (count == 0) return -1;\n"
      "  for (int device = 0; device < count; ++device)\n"
      "  {\n"
      "    cudaDeviceProp prop;\n"
      "    if (cudaSuccess == cudaGetDeviceProperties(&prop, device))\n"
      "      std::printf(\"%d.%d \", prop.major, prop.minor);\n"
      "  }\n"
      "  return 0;\n"
      "}\n")

    execute_process(COMMAND "${CUDA_NVCC_EXECUTABLE}" "--run" "${cufile}"
                    "-ccbin" ${CMAKE_CXX_COMPILER}
                    WORKING_DIRECTORY "${PROJECT_BINARY_DIR}/CMakeFiles/"
                    RESULT_VARIABLE nvcc_res OUTPUT_VARIABLE nvcc_out
                    ERROR_QUIET OUTPUT_STRIP_TRAILING_WHITESPACE)

    if(nvcc_res EQUAL 0)
      string(REPLACE "2.1" "2.1(2.0)" nvcc_out "${nvcc_out}")
      set(CUDA_GPU_DETECT_OUTPUT ${nvcc_out} CACHE INTERNAL "Returned GPU architetures from detect_gpus tool" FORCE)
    endif()
  endif()

  if(NOT CUDA_GPU_DETECT_OUTPUT)
    message(STATUS "Automatic GPU detection failed. Building for common architectures.")
    set(${OUT_VARIABLE} ${CUDA_COMMON_GPU_ARCHITECTURES} PARENT_SCOPE)
  else()
    set(${OUT_VARIABLE} ${CUDA_GPU_DETECT_OUTPUT} PARENT_SCOPE)
  endif()
endfunction()


################################################################################################
# Function for selecting GPU arch flags for nvcc based on CUDA architectures from parameter list
# Usage:
#   SELECT_NVCC_ARCH_FLAGS(out_variable [list of CUDA compute archs])
function(CUDA_SELECT_NVCC_ARCH_FLAGS out_variable)
  set(CUDA_ARCH_LIST "${ARGN}")

  if("X${CUDA_ARCH_LIST}" STREQUAL "X" )
    set(CUDA_ARCH_LIST "Auto")
  endif()

  set(cuda_arch_bin)
  set(cuda_arch_ptx)

  if("${CUDA_ARCH_LIST}" STREQUAL "All")
    set(CUDA_ARCH_LIST ${CUDA_KNOWN_GPU_ARCHITECTURES})
  elseif("${CUDA_ARCH_LIST}" STREQUAL "Common")
    set(CUDA_ARCH_LIST ${CUDA_COMMON_GPU_ARCHITECTURES})
  elseif("${CUDA_ARCH_LIST}" STREQUAL "Auto")
    CUDA_DETECT_INSTALLED_GPUS(CUDA_ARCH_LIST)
    message(STATUS "Autodetected CUDA architecture(s): ${CUDA_ARCH_LIST}")
  endif()

  # Now process the list and look for names
  string(REGEX REPLACE "[ \t]+" ";" CUDA_ARCH_LIST "${CUDA_ARCH_LIST}")
  list(REMOVE_DUPLICATES CUDA_ARCH_LIST)
  foreach(arch_name ${CUDA_ARCH_LIST})
    set(arch_bin)
    set(add_ptx FALSE)
    # Check to see if we are compiling PTX
    if(arch_name MATCHES "(.*)\\+PTX$")
      set(add_ptx TRUE)
      set(arch_name ${CMAKE_MATCH_1})
    endif()
    if(arch_name MATCHES "(^[0-9]\\.[0-9](\\([0-9]\\.[0-9]\\))?)$")
      set(arch_bin ${CMAKE_MATCH_1})
      set(arch_ptx ${arch_bin})
    else()
      # Look for it in our list of known architectures
      if(${arch_name} STREQUAL "Fermi")
        set(arch_bin "2.0 2.1(2.0)")
      elseif(${arch_name} STREQUAL "Kepler+Tegra")
        set(arch_bin 3.2)
      elseif(${arch_name} STREQUAL "Kepler+Tesla")
        set(arch_bin 3.7)
      elseif(${arch_name} STREQUAL "Kepler")
        set(arch_bin 3.0 3.5)
        set(arch_ptx 3.5)
      elseif(${arch_name} STREQUAL "Maxwell+Tegra")
        set(arch_bin 5.3)
      elseif(${arch_name} STREQUAL "Maxwell")
        set(arch_bin 5.0 5.2)
        set(arch_ptx 5.2)
      elseif(${arch_name} STREQUAL "Pascal")
        set(arch_bin 6.0 6.1)
        set(arch_ptx 6.1)
      else()
        message(SEND_ERROR "Unknown CUDA Architecture Name ${arch_name} in CUDA_SELECT_NVCC_ARCH_FLAGS")
      endif()
    endif()
    if(NOT arch_bin)
      message(SEND_ERROR "arch_bin wasn't set for some reason")
    endif()
    list(APPEND cuda_arch_bin ${arch_bin})
    if(add_ptx)
      if (NOT arch_ptx)
        set(arch_ptx ${arch_bin})
      endif()
      list(APPEND cuda_arch_ptx ${arch_ptx})
    endif()
  endforeach()

  # remove dots and convert to lists
  string(REGEX REPLACE "\\." "" cuda_arch_bin "${cuda_arch_bin}")
  string(REGEX REPLACE "\\." "" cuda_arch_ptx "${cuda_arch_ptx}")
  string(REGEX MATCHALL "[0-9()]+" cuda_arch_bin "${cuda_arch_bin}")
  string(REGEX MATCHALL "[0-9]+"   cuda_arch_ptx "${cuda_arch_ptx}")

  if(cuda_arch_bin)
    list(REMOVE_DUPLICATES cuda_arch_bin)
  endif()
  if(cuda_arch_ptx)
    list(REMOVE_DUPLICATES cuda_arch_ptx)
  endif()

  set(nvcc_flags "")
  set(nvcc_archs_readable "")

  # Tell NVCC to add binaries for the specified GPUs
  foreach(arch ${cuda_arch_bin})
    if(arch MATCHES "([0-9]+)\\(([0-9]+)\\)")
      # User explicitly specified ARCH for the concrete CODE
      list(APPEND nvcc_flags -gencode arch=compute_${CMAKE_MATCH_2},code=sm_${CMAKE_MATCH_1})
      list(APPEND nvcc_archs_readable sm_${CMAKE_MATCH_1})
    else()
      # User didn't explicitly specify ARCH for the concrete CODE, we assume ARCH=CODE
      list(APPEND nvcc_flags -gencode arch=compute_${arch},code=sm_${arch})
      list(APPEND nvcc_archs_readable sm_${arch})
    endif()
  endforeach()

  # Tell NVCC to add PTX intermediate code for the specified architectures
  foreach(arch ${cuda_arch_ptx})
    list(APPEND nvcc_flags -gencode arch=compute_${arch},code=compute_${arch})
    list(APPEND nvcc_archs_readable compute_${arch})
  endforeach()

  string(REPLACE ";" " " nvcc_archs_readable "${nvcc_archs_readable}")
  set(${out_variable}          ${nvcc_flags}          PARENT_SCOPE)
  set(${out_variable}_readable ${nvcc_archs_readable} PARENT_SCOPE)
endfunction()



================================================
FILE: exe/CMakeLists.txt
================================================
ADD_SUBDIRECTORY(luajit-rocks)


================================================
FILE: extra/CMakeLists.txt
================================================
cmake_minimum_required(VERSION 2.8)
ADD_SUBDIRECTORY(nn)
ADD_SUBDIRECTORY(graph)
ADD_SUBDIRECTORY(nngraph)
ADD_SUBDIRECTORY(nnx)
ADD_SUBDIRECTORY(threads)
ADD_SUBDIRECTORY(argcheck)

IF (WITH_CUDA STREQUAL ON)
 ADD_SUBDIRECTORY(cutorch)
 ADD_SUBDIRECTORY(cunn)
 ADD_SUBDIRECTORY(cunnx)
 ADD_SUBDIRECTORY(cudnn)
ENDIF (WITH_CUDA STREQUAL ON)



================================================
FILE: win-files/README.md
================================================
Self-contained Torch installation for windows
============

## Prerequisites

#### Must have
- MSVC, anyone of the following choices is sufficient
	- [Visual C++ Build Tools](http://landinghub.visualstudio.com/visual-cpp-build-tools)
	- [Visual Studio Express](https://www.visualstudio.com/vs/visual-studio-express/)
	- [Visual Studio Community](https://www.visualstudio.com/vs/community/)
- [Git](https://git-scm.com/download/win)
- [CMake](https://cmake.org/download/#latest)
- [Conda](http://conda.pydata.org/docs/download.html), manage dependencies like openblas, jpeg, qt, etc

#### If use CUDA

- [CUDA Toolkit](https://developer.nvidia.com/cuda-toolkit)
- [CUDNN](https://developer.nvidia.com/cudnn), if use dnn

#### Optional

- [MKL](https://software.intel.com/intel-mkl), better performance blas/lapack library
- [Gnuplot](https://sourceforge.net/projects/gnuplot/files/latest), required by gnuplot package for plotting
- [GraphicsMagick](https://sourceforge.net/projects/graphicsmagick/files/latest), optional for image package

## Install
Open "Windows Command Prompt" and run:
```bat
install.bat
```
By default x64 Torch will be installed under install\ with LuaJIT 2.1 and openlblas from conda environment 'torch-vcversion'.
Run install.bat in a specific  "VS\* \* Tools Command Prompt" to compile for a different target. Please choose x64 conda
for x64 Torch and x86 conda for x86 Torch. X86 Torch does not contain cuda packages and has 2G memory limitation.
There are a few customizable environment variables listed on top of install-deps.bat. There is no need to run
install-deps.bat before run install.bat, it sets variables in global and it will be called directly by install.bat.
**Do not** use lua instead of luajit because currently lua version Torch will use luaffifb for ffi which has bugs on windows
and has poor performance.

It is easy to intall multiple Torch by customizing TORCH\_INSTALL\_DIR, TORCH\_LUA\_VERSION, and by making sure
clean.bat is run before running install.bat.

## Use
In order to use Torch in a Self-contained way, a few helper cmd will be installed under the installation directory:
- torch-activate.cmd: setup Torch environment including TORCH\_INSTALL\_DIR, TORCH\_CONDA\_ENV, TORCH\_VS\_VERSION, TORCH\_VS\_PLATFORM, PATH, LUA\_PATH, LUA\_CPATH, CUDNN\_PATH,
- luajit.cmd: a wrapper of luajit.exe with Torch environment
- luarocks.cmd: a wrapper of luarocks.bat with Torch environment
- cmake.cmd: a wrapper of cmake.exe which helps package installation with MSVC

#### Use luajit.cmd and luarocks.cmd directly
```bat
path_to_Torch\luajit -ltorch -e "torch.test()"
```
The installation will remember which MSVC to use for what platform, so luarocks install can be run in a general "Windows
Command Prompt".
```bat
path_to_Torch\luarocks install rnn
```
It will automatically install not installed dependencies.

#### Run torch-activate.cmd, then use availabe Torch executables
```bat
path_to_Torch\torch-activate
th
qlua
```
Trepl on windows should work similarly as on linux or macos. qlua should be used to run qt related lua codes.

## Clean or Uninstall
To remove all the temporary compilation files:
```bat
clean.bat
```

To remove the installation:
```bat
path_to_Torch\torch-activate
uninstall.bat
```
torch-activate.cmd is called before uninstall.bat so that uninstall knows which Torch7 to uninstall.
In addition to clean.bat, this will remove the directory pointed to by TORCH\_INSTALL\_DIR and TORCH\_CONDA\_ENV from conda.

## Test
You can test that all libraries are installed properly by running:
```bat
path_to_Torch\torch-activate
test.bat
```
torch-activate.cmd is called before test.bat so that test knows which Torch7 to test.

Tested x64 Torch7 on Windows 10 x64, Visual Studio Community 2015, Anaconda4, Cuda Toolkit8.0, MKL2017



================================================
FILE: win-files/patch/lua-cjson.patch
================================================
diff --git a/fpconv.h b/fpconv.h
index 7b0d0ee..f674e41 100644
--- a/fpconv.h
+++ b/fpconv.h
@@ -4,6 +4,10 @@
  *
  * Longest double printed with %.14g is 21 characters long:
  * -1.7976931348623e+308 */
+#ifdef _MSC_VER
+#define inline __inline
+#define snprintf _snprintf
+#endif
 # define FPCONV_G_FMT_BUFSIZE   32
 
 #ifdef USE_INTERNAL_FPCONV
diff --git a/lua-cjson-2.1devel-1.rockspec b/lua-cjson-2.1devel-1.rockspec
index 154e333..0f7652e 100644
--- a/lua-cjson-2.1devel-1.rockspec
+++ b/lua-cjson-2.1devel-1.rockspec
@@ -47,6 +47,7 @@ build = {
     -- Override default build options (per platform)
     platforms = {
         win32 = { modules = { cjson = { defines = {
+            "strncasecmp=_strnicmp",
             "DISABLE_INVALID_NUMBERS"
         } } } }
     },
diff --git a/strbuf.c b/strbuf.c
index ac779e4..c3272f0 100644
--- a/strbuf.c
+++ b/strbuf.c
@@ -94,8 +94,8 @@ void strbuf_set_increment(strbuf_t *s, int increment)
 static inline void debug_stats(strbuf_t *s)
 {
     if (s->debug) {
-        fprintf(stderr, "strbuf(%lx) reallocs: %d, length: %d, size: %d\n",
-                (long)s, s->reallocs, s->length, s->size);
+        fprintf(stderr, "strbuf(%p) reallocs: %d, length: %d, size: %d\n",
+                s, s->reallocs, s->length, s->size);
     }
 }
 
@@ -168,8 +168,8 @@ void strbuf_resize(strbuf_t *s, int len)
     newsize = calculate_new_size(s, len);
 
     if (s->debug > 1) {
-        fprintf(stderr, "strbuf(%lx) resize: %d => %d\n",
-                (long)s, s->size, newsize);
+        fprintf(stderr, "strbuf(%p) resize: %d => %d\n",
+                s, s->size, newsize);
     }
 
     s->size = newsize;
diff --git a/strbuf.h b/strbuf.h
index d861108..b135560 100644
--- a/strbuf.h
+++ b/strbuf.h
@@ -24,6 +24,9 @@
 
 #include <stdlib.h>
 #include <stdarg.h>
+#ifdef _MSC_VER
+#define inline __inline
+#endif
 
 /* Size: Total bytes allocated to *buf
  * Length: String length, excluding optional NULL terminator.



================================================
FILE: win-files/patch/luaffifb.patch
================================================
diff --git a/ffi.h b/ffi.h
index dabdc9b..d457fd7 100644
--- a/ffi.h
+++ b/ffi.h
@@ -42,8 +42,10 @@ extern "C" {
 #include <sys/mman.h>
 #endif
 
+#ifndef _MSC_VER
 #include <complex.h>
 #define HAVE_COMPLEX
+#endif
 #define HAVE_LONG_DOUBLE
 
 #ifndef NDEBUG
diff --git a/test.c b/test.c
index 20f5d13..c9acb27 100644
--- a/test.c
+++ b/test.c
@@ -19,8 +19,12 @@
 #include <errno.h>
 #endif
 
+#ifndef _MSC_VER
 #include <complex.h>
 #define HAVE_COMPLEX
+#else
+#include <lua.h>
+#endif
 
 #ifdef __cplusplus
 # define EXTERN_C extern "C"
@@ -718,3 +722,9 @@ void test_call_pppppiiifii(void* p1, void* p2, void* p3, void* p4, void* p5, int
     sprintf(buf, "%p %p %p %p %p %d %d %d %0.1f %d %d", p1, p2, p3, p4, p5, i1, i2, i3, f4, i5, i6);
 }
 
+#ifdef _MSC_VER
+int luaopen_ffi_libtest(lua_State* L)
+{
+    return 1;
+}
+#endif
diff --git a/test.lua b/test.lua
index 8a9b718..af05ea7 100644
--- a/test.lua
+++ b/test.lua
@@ -7,6 +7,15 @@
 -- of patent rights can be found in the PATENTS file in the same directory.
 --
 io.stdout:setvbuf('no')
+
+local is_luajit = false
+if type(jit) == 'table' then
+    is_luajit = string.find(jit.version, 'LuaJIT')
+    if is_luajit then
+        package.preload['ffi'] = nil -- enable run test in luajit
+    end
+end
+
 local ffi = require 'ffi'
 local dlls = {}
 
@@ -322,17 +331,29 @@ local types = {
     e32 = 'enum e32',
 }
 
+-- different compilers may printf(%p) in different format,
+-- for example 000000003A0A9848, 0x3a0a9848
+-- tostring(cdata) may return address with '0x' prefix
+local addrbuf = ffi.new('char[256]')
+local function unifyaddr(ttype, addr)
+    if ttype == 'void*' then
+        dlls.__cdecl.print_p(addrbuf, tonumber(addr, 16))
+        return ffi.string(addrbuf)
+    end
+    return addr
+end
+
 local buf = ffi.new('char[256]')
 
 local function checkbuf(type, ret, msg)
-    local str = tostring(test_values[type]):gsub('^cdata%b<>: ', '')
+    local str = unifyaddr(type, tostring(test_values[type]):gsub('^cdata%b<>: ', ''))
     check(ffi.string(buf), str, msg)
     check(ret, #str, msg)
 end
 
 local function checkalign(type, v, ret)
     --print(v)
-    local str = tostring(test_values[type]):gsub('^cdata%b<>: ', '')
+    local str = unifyaddr(type, tostring(test_values[type]):gsub('^cdata%b<>: ', ''))
     check(ffi.string(buf), ('size %d offset %d align %d value %s'):format(ffi.sizeof(v), ffi.offsetof(v, 'v'), ffi.alignof(v, 'v'), str))
     check(ret, #str)
 end
@@ -593,9 +614,13 @@ for convention,c in pairs(dlls) do
 
     check(c.call_fptr({function(a) return 3*a end}, 5), 15)
 
+    if is_luajit then
+        print('luajit+ffifb failure : pcall error test')
+    else
     local suc, err = pcall(c.call_s, function(s) error(ffi.string(s), 0) end, 'my error')
     check(suc, false)
     check(err, 'my error')
+    end
 
     check(ffi.errno(), c.get_errno())
     c.set_errno(3)
@@ -628,7 +653,7 @@ for convention,c in pairs(dlls) do
     first = false
 end
 
-local c = ffi.C
+local c = ffi.os=='Windows' and ffi.load('msvcrt') or ffi.C
 
 assert(c.sprintf(buf, "%g", 5.3) == 3 and ffi.string(buf) == '5.3')
 assert(c.sprintf(buf, "%d", false) == 1 and ffi.string(buf) == '0')
@@ -927,31 +952,40 @@ void test_call_pppppiifiii(void* p1, void* p2, void* p3, void* p4, void* p5, int
 void test_call_pppppiiifii(void* p1, void* p2, void* p3, void* p4, void* p5, int i1, int i2, int i3, float i4, int i5, int i6);
 ]]
 
-ffi.C.test_call_echo("input")
-assert(ffi.C.buf == "input")
+local ffiC = ffi.os=='Windows' and dlls.__cdecl or ffi.C
+ffiC.test_call_echo("input")
+assert(ffiC.buf == "input")
 
 local function ptr(x) return ffi.new('void*', x) end
+local prefix = ""
+for i=1,5 do prefix = prefix .. unifyaddr('void*', tostring(i)) .. " " end
 
-ffi.C.test_call_pppppii(ptr(1), ptr(2), ptr(3), ptr(4), ptr(5), 6, 7)
-assert(ffi.C.buf == "0x1 0x2 0x3 0x4 0x5 6 7")
+ffiC.test_call_pppppii(ptr(1), ptr(2), ptr(3), ptr(4), ptr(5), 6, 7)
+assert(ffiC.buf == prefix .. "6 7")
 
-ffi.C.test_call_pppppiiiiii(ptr(1), ptr(2), ptr(3), ptr(4), ptr(5), 6, 7, 8, 9, 10, 11)
-assert(ffi.C.buf == "0x1 0x2 0x3 0x4 0x5 6 7 8 9 10 11")
+ffiC.test_call_pppppiiiiii(ptr(1), ptr(2), ptr(3), ptr(4), ptr(5), 6, 7, 8, 9, 10, 11)
+assert(ffiC.buf == prefix .. "6 7 8 9 10 11")
 
-ffi.C.test_call_pppppffffff(ptr(1), ptr(2), ptr(3), ptr(4), ptr(5), 6.5, 7.5, 8.5, 9.5, 10.5, 11.5)
-assert(ffi.C.buf == "0x1 0x2 0x3 0x4 0x5 6.5 7.5 8.5 9.5 10.5 11.5")
+if is_luajit then
+    print("luajit+ffifb failure : function call with float parameters test")
+else
+ffiC.test_call_pppppffffff(ptr(1), ptr(2), ptr(3), ptr(4), ptr(5), 6.5, 7.5, 8.5, 9.5, 10.5, 11.5)
+assert(ffiC.buf == prefix .. "6.5 7.5 8.5 9.5 10.5 11.5")
 
-ffi.C.test_call_pppppiifiii(ptr(1), ptr(2), ptr(3), ptr(4), ptr(5), 6, 7, 8.5, 9, 10, 11)
-assert(ffi.C.buf == "0x1 0x2 0x3 0x4 0x5 6 7 8.5 9 10 11")
+ffiC.test_call_pppppiifiii(ptr(1), ptr(2), ptr(3), ptr(4), ptr(5), 6, 7, 8.5, 9, 10, 11)
+assert(ffiC.buf == prefix .. "6 7 8.5 9 10 11")
 
-ffi.C.test_call_pppppiiifii(ptr(1), ptr(2), ptr(3), ptr(4), ptr(5), 6, 7, 8, 9.5, 10, 11)
-assert(ffi.C.buf == "0x1 0x2 0x3 0x4 0x5 6 7 8 9.5 10 11")
+ffiC.test_call_pppppiiifii(ptr(1), ptr(2), ptr(3), ptr(4), ptr(5), 6, 7, 8, 9.5, 10, 11)
+assert(ffiC.buf == prefix .. "6 7 8 9.5 10 11")
+end
 
-local sum = ffi.C.add_dc(ffi.new('complex', 1, 2), ffi.new('complex', 3, 5))
+if not ffiC.is_msvc then
+local sum = ffiC.add_dc(ffi.new('complex', 1, 2), ffi.new('complex', 3, 5))
 assert(ffi.istype('complex', sum))
 
-sum = ffi.C.add_fc(ffi.new('complex float', 1, 2), ffi.new('complex float', 3, 5))
+sum = ffiC.add_fc(ffi.new('complex float', 1, 2), ffi.new('complex float', 3, 5))
 assert(ffi.istype('complex float', sum))
+end
 
 ffi.cdef [[
 struct Arrays {
@@ -988,6 +1022,9 @@ assert(ffi.string(buf, 0) == '')
 assert(ffi.string(buf, ffi.new('long long', 2)) == 'aa')
 assert(ffi.string(buf, ffi.new('int', 2)) == 'aa')
 
+if is_luajit then
+    print("luajit+ffifb failure : tmpfile test")
+else
 -- Test io.tmpfile()
 ffi.cdef [[
     int fprintf ( FILE * stream, const char * format, ... );
@@ -999,5 +1036,6 @@ f:seek("set", 0)
 local str = f:read('*l')
 assert(str == 'test: foo', str)
 f:close()
+end
 
 print('Test PASSED')



================================================
FILE: win-files/patch/torch.patch
================================================
diff --git a/rocks/torch-scm-1.rockspec b/rocks/torch-scm-1.rockspec
index 2228726..451f1aa 100644
--- a/rocks/torch-scm-1.rockspec
+++ b/rocks/torch-scm-1.rockspec
@@ -27,7 +27,7 @@ cmake -E make_directory build && cd build && cmake .. -DCMAKE_BUILD_TYPE=Release
 	 platforms = {
       windows = {
            build_command = [[
-cmake -E make_directory build && cd build && cmake .. -DCMAKE_BUILD_TYPE=Release -DLUA=$(LUA) -DLUALIB=$(LUALIB) -DLUA_BINDIR="$(LUA_BINDIR)" -DLUA_INCDIR="$(LUA_INCDIR)" -DLUA_LIBDIR="$(LUA_LIBDIR)" -DLUADIR="$(LUADIR)" -DLIBDIR="$(LIBDIR)" -DCMAKE_INSTALL_PREFIX="$(PREFIX)" && $(MAKE)
+cmake -E make_directory build && cd build && cmake .. -DINTEL_MKL_DIR="$(INTEL_MKL_DIR)" -DINTEL_COMPILER_DIR="$(INTEL_COMPILER_DIR)" -DBLAS_LIBRARIES="$(BLAS_LIBRARIES)" -DLAPACK_LIBRARIES="$(LAPACK_LIBRARIES)" -DLAPACK_FOUND=$(LAPACK_FOUND) -DCMAKE_BUILD_TYPE=Release -DLUA=$(LUA) -DLUALIB=$(LUALIB) -DLUA_BINDIR="$(LUA_BINDIR)" -DLUA_INCDIR="$(LUA_INCDIR)" -DLUA_LIBDIR="$(LUA_LIBDIR)" -DLUADIR="$(LUADIR)" -DLIBDIR="$(LIBDIR)" -DCMAKE_INSTALL_PREFIX="$(PREFIX)" && $(MAKE)
 ]]
       }
    },



================================================
FILE: win-files/test/ffifb_perf_ffi.lua
================================================
-- code from http://luajit.org/ext_ffi.html

require 'torch'

local t = torch.Timer()
t:reset()

local ffi = require("ffi")
ffi.cdef[[
typedef struct { uint8_t red, green, blue, alpha; } rgba_pixel;
]]

local function image_ramp_green(n)
  local img = ffi.new("rgba_pixel[?]", n)
  local f = 255/(n-1)
  for i=0,n-1 do
    img[i].green = i*f
    img[i].alpha = 255
  end
  return img
end

local function image_to_grey(img, n)
  for i=0,n-1 do
    local y = 0.3*img[i].red + 0.59*img[i].green + 0.11*img[i].blue
    img[i].red = y; img[i].green = y; img[i].blue = y
  end
end

local N = 400*400
local img = image_ramp_green(N)
for i=1,1000 do
  image_to_grey(img, N)
end

t:stop()
print(t:time().real)



================================================
FILE: win-files/test/ffifb_perf_purelua.lua
================================================
-- code from http://luajit.org/ext_ffi.html

require 'torch'

local t = torch.Timer()
t:reset()

local floor = math.floor

local function image_ramp_green(n)
  local img = {}
  local f = 255/(n-1)
  for i=1,n do
    img[i] = { red = 0, green = floor((i-1)*f), blue = 0, alpha = 255 }
  end
  return img
end

local function image_to_grey(img, n)
  for i=1,n do
    local y = floor(0.3*img[i].red + 0.59*img[i].green + 0.11*img[i].blue)
    img[i].red = y; img[i].green = y; img[i].blue = y
  end
end

local N = 400*400
local img = image_ramp_green(N)
for i=1,1000 do
  image_to_grey(img, N)
end

t:stop()
print(t:time().real)