From 4234a8344d3204d3227a4dd69512d58714074541 Mon Sep 17 00:00:00 2001
From: Alex Hocks <alex.hocks@tum.de>
Date: Mon, 8 Jun 2026 17:50:18 +0200
Subject: [PATCH] add doc for workers

---
 docs/configuration.md | 32 ++++++++++++++++++++++++++++++++
 1 file changed, 32 insertions(+)

diff --git a/docs/configuration.md b/docs/configuration.md
index 8c44aa49..1980bb55 100644
--- a/docs/configuration.md
+++ b/docs/configuration.md
@@ -260,6 +260,38 @@ The following parameters can be set
 }
 ```
 
+## Tasking / Workers
+
+The usage of workers can be enabled by configuring the tasking parameters to values other than the default.
+When using workers, a separate MPI process group is started per Micro Manager rank. This should allow for parallel
+micro simulation computation.
+With `backend`, the communication between the Micro Manager rank and its workers can be set to use either a
+socket based or MPI based approach. For system with SLURM: By setting `is_slurm` to true the worker process is
+attempted to be started with `srun` instead of `mpiexec`. In this case, `backend` must be set to `"socket"`.
+`num_workers` controls the amount of workers per Micro Manager rank. When `num_workers` is less or equal than `1`, then
+local execution is assumed, thus no worker processes are created.
+`mpi_impl` should be set to the underlying MPI implementation, as this is required for pinning.
+
+| Parameter       | Description                                            | Default    |
+|-----------------|--------------------------------------------------------|------------|
+| `backend`       | Communication backend. Options: ["socket", "mpi"]      | `"socket"` |
+| `is_slurm`      | Launch worker with `srun`?                             | `"False"`  |
+| `"num_workers"` | Number of workers pre Micro Manager rank.              | `1`        |
+| `"mpi_impl"`    | Implementation type of MPI. Options: ["intel", "open"] | `"open"`   |
+
+The following configuration block should be provided on the same level as the `"simulation_param"`.
+The given example uses socket based communication, launches with `mpiexec`, uses 4 workers per rank and
+assumes Intel MPI.
+
+```json
+"tasking": {
+        "backend": "socket",
+        "is_slurm": false,
+        "num_workers": 4,
+        "mpi_impl": "intel"
+    }
+```
+
 ## Interpolate a crashed micro simulation
 
 If the optional dependency `sklearn` is installed, the Micro Manager will derive the output of a crashed micro simulation by interpolating outputs from similar simulations. To enable this, set