Using and Installing CCL on NERSC

To use the central installation of pyCCL at NERSC.

• Using the CCL Python module in the shared environment :
  • Setup the environment: source /global/common/software/lsst/common/miniconda/setup_current_python.sh
  • Quick test to be sure things are set up correctly You can check whether pyccl has been set up correctly by running python -c "import pyccl" and checking that no errors are returned.

or

• Run Example Notebooks

  • Set up your NERSC account by adding DESC jupyter kernels to your environment. This only needs to be done once.

  source /global/common/software/lsst/common/miniconda/kernels/setup.sh

  • Clone the CCL repo to your NERSC $HOME directory

  git clone https://github.com/LSSTDESC/CCL.git

  • Start up jupyter-dev in your web browser.

    1. Choose either

       • The simple notebook editor: http://jupyter-dev.nersc.gov/

         OR

       • The more full-featured JupyterLab available by pointing your browser to:

         https://jupyter-dev.nersc.gov/user/$USER/lab/tree$HOME

         Replace $USER with your NERSC username, and $HOME with the full path to your NERSC $HOME directory (e.g /global/homes/p/pjm43)

    2. Open your notebook

    3. Change the kernel to desc-stack

    4. Run your notebook

To Build the CCL Paper

• git clone https://github.com/LSSTDESC/CCL.git
• cd CCL
• git checkout ccl_paper
• module load python/3.6-anaconda-4.4
• module load latex
• pip install --user mkauthlist
• export MKAUTHLIST=$HOME/.local/cori/3.6-anaconda-4.4/bin/mkauthlist
• cd doc/ccl_paper
• make apj

C Library Installation

[OUTDATED] These are old notes for building the C library. Leaving this here for now.

0. If class is not installed already (The last step needs to be done in CCL installation directory).

   • export CFLAGS=-fPIC
   • export CRAYPE_LINK_TYPE=dynamic
   • export XTPE_LINK_TYPE=dynamic
   • python class_install.py

1. Load the necessary modules. You can run the steps as follows (or put them in a script, e.g., CCL_setup.sh, and do source CCL_setup.sh)

   • module load gsl/2.1
   • module load cray-fftw
   • module load swig
   • export LDFLAGS+="-L$GSL_DIR/lib -L$FFTW_DIR"
   • export CPPFLAGS+="-I$GSL_DIR/include -I$FFTW_DIR/../include"

2. Then you can go to the CCL directory and install CCL in /path/you/choose by running

   • ./configure --prefix=/path/you/choose
   • make
   • make install

3. Next, do the following to allow your compiled code to call CCL (only need to do once, or write to your .bashrc file):

   • export LD_LIBRARY_PATH=/path/you/choose/lib:$LD_LIBRARY_PATH.

   and add /path/you/choose to LDFLAGS and CPPFLAGS by (pay attention to the space before -L)

   • export LDFLAGS+=" -L/path/you/choose/lib"
   • export CPPFLAGS+=" -I/path/you/choose/include"

4. Now you can compile your code calling CCL, e.g., ./examples/ccl_sample_run.c by

   • gcc -Wall -Wpedantic -g $CPPFLAGS -std=gnu99 -fPIC examples/ccl_sample_run.c -o examples/ccl_sample_run $LDFLAGS -lgsl -lgslcblas -lm -lccl

   and run it.