@@ -17,6 +17,11 @@ The simulation can be started by the example start file:
``source start_FDSimulation.sh``
Switch of Lama warnings and Tracing
`export SCAI_UNSUPPORTED=IGNORE`
` export SCAI_TRACE=OFF`
Different kinds of parallelization are possible:
- Using OpenMP (shared-memory parallel):
...
...
@@ -26,6 +31,12 @@ Different kinds of parallelization are possible:
The standard configuration of `start_FDSimulation.sh` is using both kinds of parallelization.
Run Code on GPUs
`export SCAI_ASYNCHRONOUS=2`
`export SCAI_DEVICE=0,1,2,3 (...)` (Comma seperated list of GPU devices on one node
`export SCAI_CONTEXT=CUDA`
## Run the tests
To test the proper functionality of the installation, you can run the build in unit and integration tests.
Since the [Google Test framework](https://github.com/google/googletest) is used for the unit tests, the environment variable `GTEST_DIR` has to be set to the location of the compiled Google Test library (`libgtest.*` and `libgtest_main.*`). The integration test has no special requirements apart the LAMA framework and MPI.
...
...
@@ -34,12 +45,5 @@ Since the [Google Test framework](https://github.com/google/googletest) is used
If one of the test cases fails, the script will give an error message and will abort.
## Data processing for parallel I/O
Since the modelling code supports parallel I/O operations some pre- and post-processing steps maybe required to split or merge data. In order to use the binaries mentioned below, you have to build [LAMA](www.libama.org) with the option `BUILD_EXAMPLES=ON`.
#### Data pre-processing
- to partition a single vector-file `${SCAI_ROOT}/lama/examples/io/vectorRepartition.exe filename.mtx 1 filename.%r.mtx {NProcessors}`
#### Data post-processing
- Merge to a single vector-file `${SCAI_ROOT}/lama/examples/io/vectorRepartition.exe filename.%r.mtx {NProcessors} filename.mtx 1`
# Parallel IO
specify .lmf or .su as file type in the configuration file to use parallel in and output.
