update heisenbug doc (#651)

Documentation/gpus.rst

@@ -50,7 +50,9 @@ does not use a workload manager but is equipped with multiple GPUs per node.
 Supported SeisSol Features
 ~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-- elastic wave propagation model with kinematic point sources
+- elastic wave propagation model 
+- kinematic point sources
+- dynamic source (only for the linear slip weakening friction law)
 - off-fault plasticity model
 
 
@@ -106,44 +108,43 @@ It also has 2 GPGPUs (NVIDIA GeForce RTX 3090), that can be used to run the GPU
 The RTX 3090 belongs to a consumer kind of graphics cards and thus does not perform well with double precision. 
 Therefore, it is preferable to compile SeisSol with single precision.
 
-
-A CUDA-Aware MPI on heisenbug is loaded with:
+A module integrating all libraries relevant for SeisSol-GPU is preinstalled at ``/export/dump/ravil/modulefiles``.
+It can be discovered at startup after adding the following to ``~/.bashrc``:
 
 .. code-block:: bash
 
-    module load gcc/10.2.0
-    module load mpi.ompi/4.1.0
+    module_hpcsdk=/export/dump/ravil/modulefiles
+    export MODULEPATH=$MODULEPATH:${module_hpcsdk}
 
-GemmForge and ChainForge (for GPU) can be installed with:
+It is then loaded with:
 
 .. code-block:: bash
 
-    pip3 install git+https://github.com/ravil-mobile/gemmforge.git
-    pip3 install git+https://github.com/ravil-mobile/chainforge.git
+    module load seissol-env-gcc-11.1.0
 
-All dependencies required for the CPU version of SeisSol should also be installed.
-A folder containing all precompiled requirements is available at:
+Then clone the dr/sycl branch of SeisSol with:
 
 .. code-block:: bash
 
-    /export/dump/ulrich/SeisSol-GPU-build
+    git clone --branch dr/sycl https://github.com/SeisSol/SeisSol.git seissol-sycl
+    cd seissol-sycl
+    git submodule update --init --recursive
+
 
-To compile the GPU version of SeisSol on heisenbug, use the following cmake options ``-DDEVICE_ARCH=sm_86 -DHOST_ARCH=hsw``.
+To compile the GPU version of SeisSol on heisenbug, use the following cmake options ``-DDEVICE_ARCH=sm_86 -DHOST_ARCH=hsw -DDEVICE_BACKEND=cuda -DPRECISION=single``.
 Use ``-DCOMMTHREAD=ON`` for multiple GPUs, and ``-DCOMMTHREAD=OFF`` for one GPU.
 
 To run on one GPU (here with order 4, elastic), use simply:
 
 .. code-block:: bash
 
-    export OMPI_COMM_WORLD_LOCAL_RANK=0
-    export OMPI_COMM_WORLD_LOCAL_SIZE=1
-    export OMPI_MCA_btl=vader,self
-    SeisSol_Release_ssm_86_cuda_4_elastic parameters.par
+    ./launch ./SeisSol_Release_ssm_86_cuda_4_elastic ./parameters.par
+
+`launch` is a simple bash helper script. It is generated by CMake, in the build directory).
 
 On 2 ranks, use:
 
 .. code-block:: bash
 
-    export OMPI_MCA_btl=vader,self
     mpirun -n 2 SeisSol_Release_ssm_86_cuda_4_elastic parameters.par