jmmcg::LIBJMMCG_VER_NAMESPACE::ppd::numa_cpu_traits Struct Reference

#include <numa_traits.hpp>

Public Types
enum	: std::size_t { total_cores =12 }
enum	: bool { is_numa =true }
using	element_type = std::array< std::array< unsigned short, 6 >, 2 >

Static Public Attributes
static constexpr element_type	node_mapping {{{0, 1, 2, 3, 4, 5}, {6, 7, 8, 9, 10, 11}}}
	This provides a mapping of all sequentially-listed NUMA-nodes to the cores they contain and the mapping of those cores to to that returned by numactl and /proc/cpuinfo. More...

Detailed Description

Definition at line 34 of file numa_traits.hpp.

Member Typedef Documentation

element_type

using jmmcg::LIBJMMCG_VER_NAMESPACE::ppd::numa_cpu_traits::element_type = std::array<std::array<unsigned short, 6>, 2>

Definition at line 41 of file numa_traits.hpp.

Member Enumeration Documentation

anonymous enum

anonymous enum : std::size_t

Enumerator
total_cores

Definition at line 35 of file numa_traits.hpp.

anonymous enum

anonymous enum : bool

Enumerator
is_numa

Definition at line 38 of file numa_traits.hpp.

Member Data Documentation

node_mapping

constexpr element_type jmmcg::LIBJMMCG_VER_NAMESPACE::ppd::numa_cpu_traits::node_mapping {{{0, 1, 2, 3, 4, 5}, {6, 7, 8, 9, 10, 11}}}

inlinestaticconstexpr

This provides a mapping of all sequentially-listed NUMA-nodes to the cores they contain and the mapping of those cores to to that returned by numactl and /proc/cpuinfo.

So one can use this mapping to pin, set priority, etc physical OS-threads and assign tasks to them as appropriate to make more efficient use of the NUMA topology of the target system. Note that kernel options such as CONFIG_NUMA, CONFIG_NUMA_BALANCING, CONFIG_NUMA_BALANCING_DEFAULT_ENABLED, etc may need to be enabled.

Definition at line 47 of file numa_traits.hpp.

---

The documentation for this struct was generated from the following file:

• unix/numa_traits.hpp