GPU version of HemePure.
The user can either:
- Run the Full build script (FullBuild.sh), which first builds the dependencies and then compiles the source code or
- Do these steps as listed here:
Build dependencies before attempting to build hemepure_gpu
.
- Create
dep/build/
. - In
dep/build/
runccmake -B. -H../
orccmake ..
. - Configure using CMake.
- Run
make
indep/build/
.
- Create
src/build/
. - In
src/build/
runccmake -B. -H../
orccmake ..
. - Configure using CMake.
- Run
make
insrc/build/
.
IMPORTANT NOTE
Remember to modify the following (in the script FullBuild.sh
, or in src/CMakeLists.txt
) depending on the GPU compute capability (-gencode arch=compute_70,code=sm_70
) of the NVIDIA GPUs available on the system
-DCMAKE_CUDA_FLAGS="-ccbin g++ -gencode arch=compute_70,code=sm_70 -lineinfo --ptxas-options=-v --disable-warnings"
Running the executable can be done in the same way as the CPU version.
Launch a HemeLB simulation with the instruction:
mpirun -n N -in <input file *.xml address> -out
e.g. mpirun -n 4 ./hemepure_gpu -in ./input.xml -out results
A detailed description of the input file and how to run a simulation is provided in the official HemeLB website
http://hemelb.org/tutorials/simulation/
The folder cases/bifurcation_hires/
contains:
- the geometry input file (
bifurcation.gmy
) - the input file for the simulation (
input.xml
)
The analysis of results can be done following the instructions provided in the official HemeLB website
http://hemelb.org/tutorials/simulation/sim_section3/