Github user jpeach commented on a diff in the pull request: https://github.com/apache/trafficserver/pull/956#discussion_r77443787 --- Diff: iocore/eventsystem/UnixEventProcessor.cc --- @@ -152,6 +155,59 @@ EventProcessor::start(int n_event_threads, size_t stacksize) #else Debug("iocore_thread", "EThread: %d %s: %d", i, obj_name, obj->logical_index); #endif // HWLOC_API_VERSION + } +#endif // TS_USE_HWLOC + + snprintf(thr_name, MAX_THREAD_NAME_LENGTH, "[ET_NET %d]", i); +#if TS_USE_HWLOC + if (obj_count > 0) { + hwloc_membind_policy_t mem_policy = HWLOC_MEMBIND_DEFAULT; + hwloc_nodeset_t nodeset = hwloc_bitmap_alloc(); + int num_nodes = 0; + + hwloc_cpuset_to_nodeset(ink_get_topology(), obj->cpuset, nodeset); + num_nodes = hwloc_get_nbobjs_inside_cpuset_by_type(ink_get_topology(), obj->cpuset, HWLOC_OBJ_NODE); + + if (num_nodes == 1) { + mem_policy = HWLOC_MEMBIND_BIND; + } else if (num_nodes > 1) { + mem_policy = HWLOC_MEMBIND_INTERLEAVE; + } + + if (mem_policy != HWLOC_MEMBIND_DEFAULT) { + hwloc_set_membind_nodeset(ink_get_topology(), nodeset, mem_policy, HWLOC_MEMBIND_THREAD); + } --- End diff -- If I understand this correctly, we swizzle the memory allocation policy for the thread so that the subsequent allocations for the stacks will be interleaved across nodes. Does this depend on the mechanics of how ``malloc(3)`` chooses to allocate? Why do we have to figure this out for each thread? Can we just do it once then allocate all the stacks? Why do we even bother with ``HWLOC_MEMBIND_BIND``? If there's only 1 node, it seems like it would have to bind memory to that node?
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. ---