Merge pull request #318 from mir-group/install/yu

Compilers and libraries from conda instead of module load

CMakeLists.txt

@@ -94,7 +94,7 @@ ExternalProject_Add(
     INSTALL_COMMAND ""
 )
 if (NOT FLARE_PYTHON_VERSION)
-    set(Python_ADDITIONAL_VERSIONS 3.7 3.6 3.5)
+    set(Python_ADDITIONAL_VERSIONS 3.8 3.7 3.6 3.5)
 endif()
 find_package(PythonLibsNew ${FLARE_PYTHON_VERSION} REQUIRED)
 
@@ -157,6 +157,12 @@ add_dependencies(flare eigen_project)
 # Link to json.
 target_link_libraries(flare PUBLIC nlohmann_json::nlohmann_json)
 
+# Add conda include directories
+if (DEFINED ENV{CONDA_PREFIX})
+  message(STATUS "Adding conda include directories.")
+  include_directories($ENV{CONDA_PREFIX}/include)
+endif()
+
 # Link to OpenMP.
 if (DEFINED ENV{OMP_LOC})
     target_link_libraries(flare PUBLIC $ENV{OMP_LOC})

README.md

@@ -53,33 +53,17 @@ All the tutorials take a few minutes to run on a normal desktop computer or lapt
 
 ## Installation
 ### Pip installation
-If you're installing on a compute cluster, make sure to load the following modules first:
-```
-module load cmake/3.17.3-fasrc01 python/3.6.3-fasrc01 gcc/9.3.0-fasrc01
-```
-
-FLARE can be installed in two different ways.
-1. Download and install automatically:
-    ```
-    pip install mir-flare
-    ```
-2. Download this repository and install (required for unit tests):
-    ```
-    git clone https://github.com/mir-group/flare
-    cd flare
-    pip install .
-    ```
-    
+Please check the [installation guide here](https://mir-group.github.io/flare/installation/install.html).
 This will take a few minutes on a normal desktop computer or laptop.
 
 ### Developer's installation guide
-For developers, please check the [installation guide](https://mir-group.github.io/flare_pp/installation.html).
+For developers, please check the [installation guide](https://mir-group.github.io/flare/installation/install.html#developer-s-installation-guide).
 
 ### Compiling LAMMPS
 See [documentation on compiling LAMMPS with FLARE](https://mir-group.github.io/flare/installation/lammps.html)
 
 ### Trouble shooting
-If you have problem compiling and installing the code, please check the [FAQs](https://mir-group.github.io/flare_pp/faqs.html) to see if your problem is covered. Otherwise, please open an issue or contact us.
+If you have problem compiling and installing the code, please check the [FAQs](https://mir-group.github.io/flare/installation/install.html#trouble-shooting) to see if your problem is covered. Otherwise, please open an issue or contact us.
 
 ## System requirements
 ### Software dependencies

docs/source/installation/install.rst

@@ -5,29 +5,41 @@ Installation of FLARE
 Requirements
 ************
 
-If you're installing on a compute cluster, make sure to load the following modules first:
+1. Create a new conda environment to install flare
 
 .. code-block:: bash
 
-    module load cmake/3.17.3-fasrc01 python/3.6.3-fasrc01 gcc/9.3.0-fasrc01
+    conda create --name flare python=3.8
+    conda activate flare
 
-**********************
-Installation using pip
-**********************
+2. Use conda to install compilers and dependencies for flare
 
-Pip can automatically fetch the source code from PyPI_ and install.
+* Option 1: If you want to install flare with mkl
+
+.. code-block:: bash
 
+    conda install -y gcc gxx cmake mkl-devel mkl-service openmp -c conda-forge
+    
+* Option 2: If you want to install flare with openblas + lapacke
+
 .. code-block:: bash
 
-  $ pip install mir-flare
+    conda install -y gcc gxx cmake openmp liblapacke openblas -c conda-forge
 
-For non-admin users
+* Option 3: You can load modules if your machine have already installed them
 
 .. code-block:: bash
 
-  $ pip install --upgrade --user mir-flare
-    
-.. _PyPI: https://pypi.org/project/mir-flare/
+    module load cmake/3.17.3 gcc/9.3.0 intel-mkl/2017.2.174
+
+3. Download flare code from github repo and pip install 
+
+.. code-block:: bash
+
+    git clone -b development https://github.com/mir-group/flare.git
+    cd flare
+    pip install .
+
 
 ******************************
 Developer's installation guide
@@ -60,6 +72,42 @@ Finally, add the path of ``flare`` to ``PYTHONPATH``, such that you can ``import
 An alternative way is setting ``sys.path.append(<flare path>)`` in your python script.
 
 
+***********************
+Test FLARE installation
+***********************
+
+After the installation is done, you can leave the current folder and in the python console, try
+
+.. code-block:: ipython
+
+    >>> import flare
+    >>> flare.__file__
+    '/xxx/.conda/envs/flare/lib/python3.8/site-packages/flare/__init__.py'
+    >>> import flare.bffs.sgp
+
+You can also check that the flare C++ library is linked to mkl (or openblas and lapacke) and openmp by
+
+.. code-block:: bash
+
+    ldd /xxx/.conda/envs/flare/lib/python3.8/site-packages/flare/bffs/sgp/_C_flare.*.so
+
+where it is expected to show libmkl (or libopenblas), libgomp etc.
+
+****************
+Trouble shooting
+****************
+
+* If it fails to build mir-flare during pip install, check whether you have done `conda install cxx-compiler`, then it sometimes does not find the correct g++, i.e. `which gcc` gives the conda’s gcc, while `which g++` still gives the local one. In such case, try `conda uninstall cxx-compiler gxx`, and then redo `conda install gxx -c conda-forge`
+
+* If you get the error that `mkl.h` is not found during pip install, first check `conda list` that `mkl-include` is installed in the current environment. You can also use your own mkl headers by setting environment variable `MKL_INCLUDE` to the directory
+
+* If you manage to install flare, but get warning when `import flare.bffs.sgp`, then please check
+
+    * `which pip` should show that `pip` is in the `.conda/envs/flare/bin/` directory, instead of others
+
+    * the `ldd` command above should show the linked libraries in the `.conda/envs/flare` directory
+
+
 *****************************************
 Acceleration with multiprocessing and MKL
 *****************************************

requirements.txt

@@ -1,4 +1,4 @@
-numpy >=1.18, <=1.20.3
+numpy >=1.18, <1.23
 scipy
 memory_profiler
 numba