torch/lib/build_libs.sh

#!/usr/bin/env bash

# Shell script used to build the torch/lib/* dependencies prior to
# linking the libraries and passing the headers to the Python extension
# compilation stage. This file is used from setup.py, but can also be
# called standalone to compile the libraries outside of the overall PyTorch
# build process.

set -e

# Options for building only a subset of the libraries
WITH_CUDA=0
if [[ "$1" == "--with-cuda" ]]; then
  WITH_CUDA=1
  shift
fi

cd "$(dirname "$0")/../.."
BASE_DIR=$(pwd)
cd torch/lib
INSTALL_DIR="$(pwd)/tmp_install"
CMAKE_VERSION=${CMAKE_VERSION:="cmake"}
C_FLAGS=" -DTH_INDEX_BASE=0 -I$INSTALL_DIR/include \
  -I$INSTALL_DIR/include/TH -I$INSTALL_DIR/include/THC \
  -I$INSTALL_DIR/include/THS -I$INSTALL_DIR/include/THCS \
  -I$INSTALL_DIR/include/THNN -I$INSTALL_DIR/include/THCUNN"
# Workaround OpenMPI build failure
# ImportError: /build/pytorch-0.2.0/.pybuild/pythonX.Y_3.6/build/torch/_C.cpython-36m-x86_64-linux-gnu.so: undefined symbol: _ZN3MPI8Datatype4FreeEv
# https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=686926
C_FLAGS="${C_FLAGS} -DOMPI_SKIP_MPICXX=1"
LDFLAGS="-L$INSTALL_DIR/lib "
LD_POSTFIX=".so.1"
LD_POSTFIX_UNVERSIONED=".so"
if [[ $(uname) == 'Darwin' ]]; then
    LDFLAGS="$LDFLAGS -Wl,-rpath,@loader_path"
    LD_POSTFIX=".1.dylib"
    LD_POSTFIX_UNVERSIONED=".dylib"
else
    LDFLAGS="$LDFLAGS -Wl,-rpath,\$ORIGIN"
fi
CPP_FLAGS=" -std=c++11 "
GLOO_FLAGS=""
NCCL_ROOT_DIR=${NCCL_ROOT_DIR:-$INSTALL_DIR}
if [[ $WITH_CUDA -eq 1 ]]; then
    GLOO_FLAGS="-DUSE_CUDA=1 -DNCCL_ROOT_DIR=$NCCL_ROOT_DIR"
fi
CWRAP_FILES="\
$BASE_DIR/torch/lib/ATen/Declarations.cwrap;\
$BASE_DIR/torch/lib/ATen/Local.cwrap;\
$BASE_DIR/torch/lib/THNN/generic/THNN.h;\
$BASE_DIR/torch/lib/THCUNN/generic/THCUNN.h;\
$BASE_DIR/torch/lib/ATen/nn.yaml"
CUDA_NVCC_FLAGS=$C_FLAGS
if [[ $CUDA_DEBUG -eq 1 ]]; then
  CUDA_NVCC_FLAGS="$CUDA_NVCC_FLAGS -g -G"
fi

# Used to build an individual library, e.g. build TH
function build() {
  # We create a build directory for the library, which will
  # contain the cmake output
  mkdir -p build/$1
  cd build/$1
  BUILD_C_FLAGS=''
  case $1 in
      THCS | THCUNN ) BUILD_C_FLAGS=$C_FLAGS;;
      nanopb ) BUILD_C_FLAGS=$C_FLAGS" -fPIC -fexceptions";;
      *) BUILD_C_FLAGS=$C_FLAGS" -fexceptions";;
  esac
  ${CMAKE_VERSION} ../../$1 -DCMAKE_MODULE_PATH="$BASE_DIR/cmake/FindCUDA" \
              -DTorch_FOUND="1" \
              -DCMAKE_INSTALL_PREFIX="$INSTALL_DIR" \
              -DCMAKE_C_FLAGS="$BUILD_C_FLAGS" \
              -DCMAKE_CXX_FLAGS="$BUILD_C_FLAGS $CPP_FLAGS" \
              -DCMAKE_EXE_LINKER_FLAGS="$LDFLAGS" \
              -DCMAKE_SHARED_LINKER_FLAGS="$LDFLAGS" \
              -DCMAKE_INSTALL_LIBDIR="$INSTALL_DIR/lib" \
              -DCUDA_NVCC_FLAGS="$CUDA_NVCC_FLAGS" \
              -DCMAKE_PREFIX_PATH="$INSTALL_DIR" \
              -Dcwrap_files="$CWRAP_FILES" \
              -DTH_INCLUDE_PATH="$INSTALL_DIR/include" \
              -DTH_LIB_PATH="$INSTALL_DIR/lib" \
              -DTH_LIBRARIES="$INSTALL_DIR/lib/libTH$LD_POSTFIX" \
              -DATEN_LIBRARIES="$INSTALL_DIR/lib/libATen$LD_POSTFIX" \
              -DTHNN_LIBRARIES="$INSTALL_DIR/lib/libTHNN$LD_POSTFIX" \
              -DTHCUNN_LIBRARIES="$INSTALL_DIR/lib/libTHCUNN$LD_POSTFIX" \
              -DTHS_LIBRARIES="$INSTALL_DIR/lib/libTHS$LD_POSTFIX" \
              -DTHC_LIBRARIES="$INSTALL_DIR/lib/libTHC$LD_POSTFIX" \
              -DTHCS_LIBRARIES="$INSTALL_DIR/lib/libTHCS$LD_POSTFIX" \
              -DTH_SO_VERSION=1 \
              -DTHC_SO_VERSION=1 \
              -DTHNN_SO_VERSION=1 \
              -DTHCUNN_SO_VERSION=1 \
              -DTHD_SO_VERSION=1 \
              -DNO_CUDA=$((1-$WITH_CUDA)) \
              -DNCCL_EXTERNAL=1 \
              -Dnanopb_BUILD_GENERATOR=0 \
              -DCMAKE_DEBUG_POSTFIX="" \
              -DCMAKE_BUILD_TYPE=$([ $DEBUG ] && echo Debug || echo Release) \
              ${@:2}
  make install -j$(getconf _NPROCESSORS_ONLN)
  cd ../..

  local lib_prefix=$INSTALL_DIR/lib/lib$1
  if [ -f "$lib_prefix$LD_POSTFIX" ]; then
    rm -rf -- "$lib_prefix$LD_POSTFIX_UNVERSIONED"
  fi

  if [[ $(uname) == 'Darwin' ]]; then
    cd tmp_install/lib
    for lib in *.dylib; do
      echo "Updating install_name for $lib"
      install_name_tool -id @rpath/$lib $lib
    done
    cd ../..
  fi
}

function build_nccl() {
   mkdir -p build/nccl
   cd build/nccl
   ${CMAKE_VERSION} ../../nccl -DCMAKE_MODULE_PATH="$BASE_DIR/cmake/FindCUDA" \
               -DCMAKE_BUILD_TYPE=Release \
               -DCMAKE_INSTALL_PREFIX="$INSTALL_DIR" \
               -DCMAKE_C_FLAGS="$C_FLAGS" \
               -DCMAKE_CXX_FLAGS="$C_FLAGS $CPP_FLAGS"
   make install
   mkdir -p ${INSTALL_DIR}/lib
   cp "lib/libnccl.so.1" "${INSTALL_DIR}/lib/libnccl.so.1"
   if [ ! -f "${INSTALL_DIR}/lib/libnccl.so" ]; then
     ln -s "${INSTALL_DIR}/lib/libnccl.so.1" "${INSTALL_DIR}/lib/libnccl.so"
   fi
   cd ../..
}

# purpusefully not using build() because we need ATen to build the same
# regardless of whether it is inside pytorch or not, so it
# cannot take any special flags
# special flags need to be part of the ATen build itself
#
# However, we do explicitly pass library paths when setup.py has already
# detected them (to ensure that we have a consistent view between the
# PyTorch and ATen builds.)
function build_aten() {
  mkdir -p build/aten
  cd  build/aten
  ${CMAKE_VERSION} ../../../../aten \
  -DCMAKE_BUILD_TYPE=$([ $DEBUG ] && echo Debug || echo Release) \
  -DNO_CUDA=$((1-$WITH_CUDA)) \
  -DCUDNN_INCLUDE_DIR=$CUDNN_INCLUDE_DIR \
  -DCUDNN_LIB_DIR=$CUDNN_LIB_DIR \
  -DATEN_NO_CONTRIB=1 \
  -DCMAKE_INSTALL_PREFIX="$INSTALL_DIR"
  # purpusefully not passing C_FLAGS for the same reason as above
  make -j$(getconf _NPROCESSORS_ONLN) install
  cd ../..
}

# In the torch/lib directory, create an installation directory
mkdir -p tmp_install

# Build
for arg in "$@"; do
    if [[ "$arg" == "nccl" ]]; then
        build_nccl
    elif [[ "$arg" == "gloo" ]]; then
        build gloo $GLOO_FLAGS
    elif [[ "$arg" == "ATen" ]]; then
        build_aten
    else
        build $arg
    fi
done

# If all the builds succeed we copy the libraries, headers,
# binaries to torch/lib
rm -rf $INSTALL_DIR/lib/cmake
rm -rf $INSTALL_DIR/lib/python
cp $INSTALL_DIR/lib/* .
if [ -d "$INSTALL_DIR/lib64/" ]; then
    cp $INSTALL_DIR/lib64/* .
fi
cp ../../aten/src/THNN/generic/THNN.h .
cp ../../aten/src/THCUNN/generic/THCUNN.h .
cp -r $INSTALL_DIR/include .
if [ -d "$INSTALL_DIR/bin/" ]; then
    cp $INSTALL_DIR/bin/* .
fi

# this is for binary builds
if [[ $PYTORCH_BINARY_BUILD && $PYTORCH_SO_DEPS ]]
then
    echo "Copying over dependency libraries $PYTORCH_SO_DEPS"
    # copy over dependency libraries into the current dir
    cp $PYTORCH_SO_DEPS .
fi