The following list will be amended and updated regularly. The real and up-to-date list of compilers (and versions) available can be seen with the command module avail
or module spider
, respectively.
Currently (April 2020) installed are:
| CUDA | https://developer.nvidia.com/cuda-zone/ |
| GCC | https://gcc.gnu.org/ |
| Intel | https://software.intel.com/en-us/parallel-studio-xe |
| IntelMPI | https://software.intel.com/en-us/mpi-library |
| Java | https://openjdk.java.net/ |
| LLVM | https://llvm.org/ |
| OpenMPI | https://www.open-mpi.org/ |
|
Python (incl. numpy, scipy, matplotlib, ipython, jupyter, pandas, h5py, keras, tensorflow, torch) |
https://www.python.org/ |
| R | https://www.r-project.org/ |
Compiler options
There are various options to achieve the best performance with a given compiler.
We recommend to set the following parameters:
Intel Compiler: '-O2 -march=sandybridge -mtune=haswell'
GNU Compiler: '-O2 -march=sandybridge -mtune=haswell'
The options '-Ofast' or '-O3' are possible, too. However, they are not suitable to all codes, and can lead to problems with some programs.
A few older compiler versions may require the options -march=corei7-avx -mtune=core-avx2 instead.