Build-instructions.md

@@ -25,8 +25,8 @@ the darts python package directory, for example `WPy64-39100\python-3.9.10.amd64
 ## Linux (multithread)
 
 1. Install python3, python3-dev, and pip
-2. Install numpy (`pip install numpy`)
-3. Run `build_darts.sh 0` for build multithread engines with iterative linear solvers or `build_darts.sh 1` for build sequential engines with ODLS. Pass the second argument to use more cores for build, for example with 8 threads: `build_darts.sh 0 8`.
+2. Run `build_darts.sh 0` for build multithread engines with iterative linear solvers or `build_darts.sh 1` for build sequential engines with ODLS. Pass the second argument to use more cores for build, for example with 8 threads: `build_darts.sh 0 8`.
+3. Install the python package: `pip install .`
 
 ## Linux (multithread+GPU)
 
@@ -48,3 +48,14 @@ use this [solution](https://github.com/NVIDIA/nccl/issues/102#issuecomment-10214
 
 3. Do the steps from the Linux (multithread) instruction with `make gpu` (multithread with GPU support) instead of `make mt` (multithread).
 
+## Supercomputer (DelftBlue)
+1. ssh <you_net_id>@login.delftblue.tudelft.nl
+2. `cd scratch/<user_name>`
+3. clone repo: `git clone https://gitlab.com/open-darts/open-darts.git`
+4. `cd open-darts`
+5. module load 2022r2 miniconda3/4.12.0 cmake cuda/11.6 slurm
+6. activate conda environment, if you don't have one create one as follows:
+    6.1. conda create --name darts_py39 python=3.9
+    6.2. conda activate darts_py39
+7. `source build_darts.sh 1 4` This will install with open-darts-linear-solvers
+8. `pip install .`