Probably you may have already known how to install multicore xgboost using clang instead of gcc. If someone don’t know that please read on. I have a quad core laptop with High Sierra. When I install xgboost with conda or pip it runs only in one core. I read somewhere that conda installation will not be doing multicore processing. So I tried the source installation as described in the xgboost document

  1. brew install gcc@8
  2. git clone –recursive https://github.com/dmlc/xgboost
  3. mkdir build
  4. cd build
  5. CC=gcc-8 CXX=g++-8 cmake ..
  6. make -j4

However, gcc doesn’t support the option ‘-stdlib=libc++’. There is an issue with gcc which can be seen in this post (Note that if some build gcc as described in that post will be able to install xgboost without any problem). So I did the following with clang

  1. brew install llvm

    This install clang* at /usr/local/Cellar/llvm/VERSION

  2. export PATH=/usr/local/Cellar/llvm/VERSION/bin:$PATH
  3. export LDFLAGS=”-L/usr/local/Cellar/llvm/VERSION/lib”
  4. export CPPFLAGS=”-I/usr/local/Cellar/llvm/VERSION/include”
  5. git clone –recursive https://github.com/dmlc/xgboost
  6. cd xgboost
  7. mkdir build
  8. cd build

  9. CC=clang-mp-8.0 CXX=clang++-mp-8.0 cmake ..

    Note that I have replaced gcc-8 and g++-8 to clang-mp-8.0 and clang++-mp-8.0. It doesn’t work with clang-8.0 and clang++-8.0

  10. make -j4
  11. cd ../python_package
  12. python setup.py install

These steps run xgboost in multicore.