Configuration
SpecFWAT is built using CMake, a cross-platform build system. The following instructions will guide you through the process of building SpecFWAT on your system.
Create a build directory
mkdir build && cd build
configure the build
Bascially, you can use the following command to configure the build:
cmake ..
More options
1. Specify Compilers:
If you have multiple compilers installed, you can specify the Fortran and C compilers using the FC and CC options, respectively. For example on Niagara@CCDB:
CC=icc CXX=icpc FC=mpifort MPIFC=mpifort cmake ..
More options can be specified in the cmake command. For example, you can specify the HDF5 directory and enable GPU support. Here are some common options:
2. Specify the HDF5 directory:
If the hdf5 library is not installed in library path, you can specify its path using the HDF5_ROOT option. For example:
cmake -DHDF5_ROOT=/path/to/hdf5 ..
3. Enable GPU support:
Currently, SpecFWAT only supports CUDA for GPU acceleration. The CUDA architecture can be specified using the CUDA_ARCH option. Check the table below for the supported architectures:
| CUDA_ARCH | Architecture | Example Card |
|---|---|---|
| 0 (default) | GPU on current machine | N/A |
| 4 | Tesla | K10, Geforce GTX 650 |
| 5 | Kepler | K20 |
| 6 | Kepler | K80 |
| 7 | Maxwell | Quadro K2200 |
| 8 | Pascal | P100 |
| 9 | Volta | V100 |
| 10 | Turing | GeForce RTX 2080 |
| 11 | Ampere | A100 |
| 12 | Hopper | H100 |
For example, to enable CUDA support and set the architecture to Ampere (A100), you can use the following command:
cmake -DUSE_CUDA=on -DCUDA_ARCH=11 ..