Add MPI clarity to README.

README.md

@@ -177,7 +177,7 @@ cmake --build . --target all
 | `FLAMEGPU_VERBOSE_PTXAS`             | `ON`/`OFF`                  | Enable verbose PTXAS output during compilation. Default `OFF`.                                             |
 | `FLAMEGPU_CURAND_ENGINE`             | `XORWOW` / `PHILOX` / `MRG` | Select the CUDA random engine. Default `XORWOW`                                                            |
 | `FLAMEGPU_ENABLE_GLM`                | `ON`/`OFF`                  | Experimental feature for GLM type support within models. Default `OFF`.                                    |
-| `FLAMEGPU_ENABLE_MPI`                | `ON`/`OFF`                  | Enable MPI support for distributed CUDAEnsembles. Default `OFF`.                                           |
+| `FLAMEGPU_ENABLE_MPI`                | `ON`/`OFF`                  | Enable MPI support for distributed CUDAEnsembles, each MPI worker should have exclusive access to it's GPUs e.g. 1 MPI worker per node. Default `OFF`.                                           |
 | `FLAMEGPU_SHARE_USAGE_STATISTICS`    | `ON`/`OFF`                  | Share usage statistics ([telemetry](https://docs.flamegpu.com/guide/telemetry)) to support evidencing usage/impact of the software. Default `ON`. |
 | `FLAMEGPU_TELEMETRY_SUPPRESS_NOTICE` | `ON`/`OFF`                  | Suppress notice encouraging telemetry to be enabled, which is emitted once per binary execution if telemetry is disabled. Defaults to `OFF`, or the value of a system environment variable of the same name. |
 | `FLAMEGPU_TELEMETRY_TEST_MODE`       | `ON`/`OFF`                  | Submit telemetry values to the test mode of TelemetryDeck. Intended for use during development of FLAMEGPU rather than use. Defaults to `OFF`, or the value of a system environment variable of the same name.|