Deploying to gh-pages from @ 5d4993b 🚀

_sources/user-guide/XCI.rst.txt

@@ -20,7 +20,26 @@ Cray XC50 cabinet, codenamed XCI, containing 20’992 ARM cores. Available since
 Running
 =======
 
-Since the XCI system uses Cray ALPS to run jobs all compute intensive executables need to be run with ``aprun`` to run on the compute nodes otherwise the jobs can run on the MOM nodes that can cause problems.
+Since the XCI system uses Cray ALPS to run jobs all compute intensive executables need to be run with ``aprun`` to run on the compute nodes otherwise the jobs can run on the MOM nodes that can cause problems. e.g
+
+```
+aprun -n 64 my_mpi.exe
+```
+
+For multithreaded OpenMP applications you could try:
+
+```
+export OMP_NUM_THREADS=64
+aprun -d 64 my_openmp.exe
+```
+
+For hybrid multithread with 64 MPI tasks and 4 OpenMP threads (using 4 hardware threads on XCI) you could use
+
+```
+export OMP_NUM_THREADS=4
+aprun -n 64 -d 4 -j 4
+```
+
 
 Queues
 ======

user-guide/XCI.html

@@ -144,7 +144,20 @@ <h1>XCI - Marvell Thunder X2<a class="headerlink" href="#xci-marvell-thunder-x2"
 </ul>
 <section id="running">
 <h2>Running<a class="headerlink" href="#running" title="Link to this heading"></a></h2>
-<p>Since the XCI system uses Cray ALPS to run jobs all compute intensive executables need to be run with <code class="docutils literal notranslate"><span class="pre">aprun</span></code> to run on the compute nodes otherwise the jobs can run on the MOM nodes that can cause problems.</p>
+<p>Since the XCI system uses Cray ALPS to run jobs all compute intensive executables need to be run with <code class="docutils literal notranslate"><span class="pre">aprun</span></code> to run on the compute nodes otherwise the jobs can run on the MOM nodes that can cause problems. e.g</p>
+<p><code class="docutils literal notranslate"><span class="pre">`</span>
+<span class="pre">aprun</span> <span class="pre">-n</span> <span class="pre">64</span> <span class="pre">my_mpi.exe</span>
+<span class="pre">`</span></code></p>
+<p>For multithreaded OpenMP applications you could try:</p>
+<p><code class="docutils literal notranslate"><span class="pre">`</span>
+<span class="pre">export</span> <span class="pre">OMP_NUM_THREADS=64</span>
+<span class="pre">aprun</span> <span class="pre">-d</span> <span class="pre">64</span> <span class="pre">my_openmp.exe</span>
+<span class="pre">`</span></code></p>
+<p>For hybrid multithread with 64 MPI tasks and 4 OpenMP threads (using 4 hardware threads on XCI) you could use</p>
+<p><code class="docutils literal notranslate"><span class="pre">`</span>
+<span class="pre">export</span> <span class="pre">OMP_NUM_THREADS=4</span>
+<span class="pre">aprun</span> <span class="pre">-n</span> <span class="pre">64</span> <span class="pre">-d</span> <span class="pre">4</span> <span class="pre">-j</span> <span class="pre">4</span>
+<span class="pre">`</span></code></p>
 </section>
 <section id="queues">
 <h2>Queues<a class="headerlink" href="#queues" title="Link to this heading"></a></h2>