--- name: "memkindallocator" section: "3" platform: "linux" package: "libmemkind-dev" version: "1.14.0-2" release: "Debian 12 (Bookworm)" platforms: ["linux"] url: "https://new.carta.tech/memkindallocator" --- # memkindallocator(3) The C++ allocator compatible with the C++ standard library allocator concepts Note: memkind\_allocator.h functionality is considered as stable API (STANDARD API). ## NAME libmemkind::static\_kind::allocator\ - The C++ allocator compatible with the C++ standard library allocator concepts\ **Note:** *memkind\_allocator.h* functionality is considered as stable API (STANDARD API). ## SYNOPSIS ``` #include Link with -lmemkind libmemkind::static_kind::allocator(libmemkind::kinds kind); template libmemkind::static_kind::allocator::allocator(const libmemkind::static_kind::allocator&) noexcept; template libmemkind::static_kind::allocator(const allocator&& other) noexcept; libmemkind::static_kind::allocator::~allocator(); T *libmemkind::static_kind::allocator::allocate(std::size_t n) const; void libmemkind::static_kind::allocator::deallocate(T *p, std::size_t n) const; template void libmemkind::static_kind::allocator::construct(U *p, Args... args) const; void libmemkind::static_kind::allocator::destroy(T *p) const; ``` ## DESCRIPTION The **libmemkind::static\_kind::allocator\** is intended to be used with STL containers to allocate from static kinds memory. Memory management is based on memkind library. Refer [**memkind**(3)](/memkind/3.md) man page for more details. The **libmemkind::kinds** specifies allocator static kinds of memory, representing type of memory which offers different characteristics. The available types of allocator kinds of memory: **libmemkind::kinds::DEFAULT** Default allocation using standard memory and default page size. **libmemkind::kinds::HIGHEST\_CAPACITY** Allocate from a NUMA node(s) that has the highest capacity among all nodes in the system. **libmemkind::kinds::HIGHEST\_CAPACITY\_PREFERRED** Same as **libmemkind::kinds::HIGHEST\_CAPACITY** except that if there is not enough memory in the NUMA node that has the highest capacity in the local domain to satisfy the request, the allocation will fall back on other memory NUMA nodes. **Note:** For this kind, the allocation will not succeed if there are two or more NUMA nodes that have the highest capacity. **libmemkind::kinds::HIGHEST\_CAPACITY\_LOCAL** Allocate from a NUMA node that has the highest capacity among all NUMA Nodes from the local domain. NUMA Nodes have the same local domain for a set of CPUs associated with them, e.g. socket or sub-NUMA cluster. **Note:** If there are multiple NUMA nodes in the same local domain that have the highest capacity - allocation will be done from NUMA node with worse latency attribute. This kind requires locality information described in **SYSTEM CONFIGURATION** section. **libmemkind::kinds::HIGHEST\_CAPACITY\_LOCAL\_PREFERRED** Same as **libmemkind::kinds::HIGHEST\_CAPACITY\_LOCAL** except that if there is not enough memory in the NUMA node that has the highest capacity to satisfy the request, the allocation will fall back on other memory NUMA nodes. **libmemkind::kinds::LOWEST\_LATENCY\_LOCAL** Allocate from a NUMA node that has the lowest latency among all NUMA Nodes from the local domain. NUMA Nodes have the same local domain for a set of CPUs associated with them, e.g. socket or sub-NUMA cluster. **Note:** If there are multiple NUMA nodes in the same local domain that have the lowest latency - allocation will be done from NUMA node with smaller memory capacity. This kind requires locality and memory performance characteristics information described in **SYSTEM CONFIGURATION** section. **libmemkind::kinds::LOWEST\_LATENCY\_LOCAL\_PREFERRED** Same as **libmemkind::kinds::LOWEST\_LATENCY\_LOCAL** except that if there is not enough memory in the NUMA node that has the lowest latency to satisfy the request, the allocation will fall back on other memory NUMA nodes. **libmemkind::kinds::HIGHEST\_BANDWIDTH\_LOCAL** Allocate from a NUMA node that has the highest bandwidth among all NUMA Nodes from the local domain. NUMA Nodes have the same local domain for a set of CPUs associated with them, e.g. socket or sub-NUMA cluster. **Note:** If there are multiple NUMA nodes in the same local domain that have the highest bandwidth - allocation will be done from NUMA node with smaller memory capacity. This kind requires locality and memory performance characteristics information described in **SYSTEM CONFIGURATION** section. **libmemkind::kinds::HIGHEST\_BANDWIDTH\_LOCAL\_PREFERRED** Same as **libmemkind::kinds::HIGHEST\_BANDWIDTH\_LOCAL** except that if there is not enough memory in the NUMA node that has the highest bandwidth to satisfy the request, the allocation will fall back on other memory NUMA nodes. **libmemkind::kinds::HUGETLB** Allocate from standard memory using huge pages. Note: This kind requires huge pages configuration described in **SYSTEM CONFIGURATION** section. **libmemkind::kinds::INTERLEAVE** Allocate pages interleaved across all NUMA nodes with transparent huge pages disabled. **libmemkind::kinds::HBW** Allocate from the closest high bandwidth memory NUMA node at the time of allocation. If there is not enough high bandwidth memory to satisfy the request, errno is set to ENOMEM and the allocated pointer is set to *NULL*. Note: This kind requires memory performance characteristics information described in **SYSTEM CONFIGURATION** section. **libmemkind::kinds::HBW\_ALL** Same as **libmemkind::kinds::HBW** except decision regarding closest NUMA node is postponed until the time of first write. **libmemkind::kinds::HBW\_HUGETLB** Same as **libmemkind::kinds::HBW** except the allocation is backed by huge pages. Note: This kind requires huge pages configuration described in **SYSTEM CONFIGURATION** section. **libmemkind::kinds::HBW\_ALL\_HUGETLB** Combination of **libmemkind::kinds::HBW\_ALL** and **libmemkind::kinds::HBW\_HUGETLB** properties. Note: This kind requires huge pages configuration described in **SYSTEM CONFIGURATION** section. **libmemkind::kinds::HBW\_PREFERRED** Same as **libmemkind::kinds::HBW** except that if there is not enough high bandwidth memory to satisfy the request, the allocation will fall back on standard memory. **libmemkind::kinds::HBW\_PREFERRED\_HUGETLB** Same as **libmemkind::kinds::HBW\_PREFERRED** except the allocation is backed by huge pages. Note: This kind requires huge pages configuration described in **SYSTEM CONFIGURATION** section. **libmemkind::kinds::HBW\_INTERLEAVE** Same as **libmemkind::kinds::HBW** except that the pages that support the allocation are interleaved across all high bandwidth nodes and transparent huge pages are disabled. **libmemkind::kinds::REGULAR** Allocate from regular memory using the default page size. Regular means general purpose memory from the NUMA nodes containing CPUs. **libmemkind::kinds::DAX\_KMEM** Allocate from the closest persistent memory NUMA node at the time of allocation. If there is not enough memory in the closest persistent memory NUMA node to satisfy the request, *errno* is set to **ENOMEM** and the allocated pointer is set to *NULL*. **libmemkind::kinds::DAX\_KMEM\_ALL** Allocate from the closest persistent memory NUMA node available at the time of allocation. If there is not enough memory on any of persistent memory NUMA nodes to satisfy the request, *errno* is set to **ENOMEM** and the allocated pointer is set to *NULL*. **libmemkind::kinds::DAX\_KMEM\_PREFERRED** Same as **libmemkind::kinds::DAX\_KMEM** except that if there is not enough memory in the closest persistent memory NUMA node to satisfy the request, the allocation will fall back on other memory NUMA nodes. **Note:** For this kind, the allocation will not succeed if two or more persistent memory NUMA nodes are in the same shortest distance to the same CPU on which process is eligible to run. Check on that eligibility is done upon starting the application. **libmemkind::kinds::DAX\_KMEM\_INTERLEAVE** Same as **libmemkind::kinds::DAX\_KMEM** except that the pages that support the allocation are interleaved across all persistent memory NUMA nodes. All public member types and functions correspond to standard library allocator concepts and definitions. The current implementation supports C++11 standard. Template arguments:\ *T* is an object type aliased by value\_type.\ *U* is an object type. **Note:**\ **T \*libmemkind::static\_kind::allocator\::allocate(std::size\_t** *n***)** allocates memory using **memkind\_malloc**(). Throw *std::bad\_alloc* when: *n* = 0 or there is not enough memory to satisfy the request. **libmemkind::static\_kind::allocator\::deallocate(T** *\*p***, std::size\_t** *n***)** deallocates memory associated with pointer returned by **allocate**() using **memkind\_free**(). ## SYSTEM CONFIGURATION Interfaces for obtaining 2MB (HUGETLB) memory need allocated huge pages in the kernel's huge page pool. - **HUGETLB (huge pages)** Current number of "persistent" huge pages can be read from */proc/sys/vm/nr\_hugepages* file. Proposed way of setting hugepages is: **sudo sysctl vm.nr\_hugepages=\**. More information can be found here: ⟨https://www.kernel.org/doc/Documentation/vm/hugetlbpage.txt⟩ Interfaces for obtaining locality information are provided by *libhwloc* dependency. Functionality based on locality requires that memkind library is configured and built with the support of *libhwloc* (./configure --enable-hwloc). Interfaces for obtaining memory performance characteristics information are based on *HMAT* (Heterogeneous Memory Attribute Table) ⟨https://uefi.org/sites/default/files/resources/ACPI\_6\_3\_final\_Jan30.pdf⟩ Functionality based on memory performance characteristics requires that platform configuration fully supports HMAT and memkind library is configured and built with the support of *libhwloc* (./configure --enable-hwloc). **Note:** For a given target NUMA Node, the OS exposes only the performance characteristics of the best performing NUMA node. *libhwloc* can be reached on: ⟨https://www.open-mpi.org/projects/hwloc⟩ ## COPYRIGHT Copyright (C) 2019 - 2021 Intel Corporation. All rights reserved. ## SEE ALSO [**memkind**(3)](/memkind/3.md)