{"id":397814,"date":"2024-12-18T00:40:57","date_gmt":"2024-12-18T00:40:57","guid":{"rendered":"https:\/\/innovationspace.ansys.com\/forum\/forums\/reply\/397814\/"},"modified":"2024-12-18T00:40:57","modified_gmt":"2024-12-18T00:40:57","slug":"397814","status":"publish","type":"reply","link":"https:\/\/innovationspace.ansys.com\/forum\/forums\/reply\/397814\/","title":{"rendered":"Reply To: speed up the job in hpc (mpp)"},"content":{"rendered":"<p>&lt;p&gt;Hello,&lt;\/p&gt;&lt;p&gt;After you make sure you run on 32 cores, you can try to bind processes to cores and see if you can speedup the calculation further. I found the following in our knowledge database:&lt;\/p&gt;&lt;p&gt;&#8221;&lt;\/p&gt;<\/p>\n<pre>By default, an MPI process migrates between cores as the OS manages resources and attempts to\nget the best load balance on the system.\nBut because LS-DYNA is a memory intensive application, such migration can significantly degrade performance\nsince memory access can take longer if the process is moved to a core farther from the memory it is using.\nTo avoid this performance degradation, it is important to bind each MPI process to a core.  \n\nEach MPI has its own way of binding the processes to cores, and furthermore, \nthreaded MPP (HYBRID) employs a different strategy from pure MPP.\n\n\nI.  Pure MPP\n============\n\nTo bind processes to cores, include the following MPI execution line directives according to the type of MPI used.\n\nHP-MPI, Platform MPI, and IBM Platform MPI:\n\n\t-cpu_bind or -cpu_bind=rank\n        -cpu_bind=MAP_CPU:0,1,2,...   &lt;&lt;&lt;&lt; not recommended unless user really needs to bind MPI processes to specific cores\n\nIBM Platform MPI 9.1.4 and later:\n\n\t-affcycle=numa\n\nIntel MPI:\n\n\t-genv I_MPI_PIN_DOMAIN=core\n\nOpen MPI:\n\n\t--bind-to numa\n\"&lt;br&gt;&lt;br&gt;&lt;br&gt;<\/pre>\n<p>&lt;p&gt;Let me know how it goes.&lt;\/p&gt;&lt;p&gt;&nbsp;&lt;\/p&gt;&lt;p&gt;Reno.&lt;\/p&gt;<\/p>\n","protected":false},"template":"","class_list":["post-397814","reply","type-reply","status-publish","hentry"],"aioseo_notices":[],"acf":[],"_links":{"self":[{"href":"https:\/\/innovationspace.ansys.com\/forum\/wp-json\/wp\/v2\/replies\/397814","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/innovationspace.ansys.com\/forum\/wp-json\/wp\/v2\/replies"}],"about":[{"href":"https:\/\/innovationspace.ansys.com\/forum\/wp-json\/wp\/v2\/types\/reply"}],"version-history":[{"count":0,"href":"https:\/\/innovationspace.ansys.com\/forum\/wp-json\/wp\/v2\/replies\/397814\/revisions"}],"wp:attachment":[{"href":"https:\/\/innovationspace.ansys.com\/forum\/wp-json\/wp\/v2\/media?parent=397814"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}