To provide the in-memory size of Erlang terms, ignoring where the Erlang terms are stored.

Size information should match the Erlang Efficiency Guide memory information:

• Small integer: 1 word
  • On 32-bit architectures: -134217729 < i < 134217728 (28 bits)
  • On 64-bit architectures: -576460752303423489 < i < 576460752303423488 (60 bits)
• Big integer: 3..N words
• Atom: 1 word
• Float:
  • On 32-bit architectures: 4 words
  • On 64-bit architectures: 3 words
• Binary: 3..6 + data
• List: 1 word + 1 word per element + the size of each element
• Tuple: 2 words + the size of each element
• Small Map (N =< 32): 4 words + 2 words per key/value pair + the size of each pair
• Large Map (N > 32): At least, 2 words + 2 * N words + 2 * log16(N) words + the size of each pair
• Pid:
  • From local node: 1 word
  • From remote node: 5 words
• Port:
  • From local node: 1 word
  • From remote node: 5 words
• Reference:
  • On 32-bit architectures:
    • From local node: 5 words
    • From remote node: 7 words
  • On 64-bit architectures:
    • From local node: 4 words
    • From remote node: 6 words
• Fun: 9..13 words + size of environment

Maps use a "flatmap" (a pair of tuples) before switching to a Hash Array Mapped Trie (HAMT) implementation after the size becomes larger than MAP_SMALL_MAP_LIMIT (32). The upper-limit of a HAMT map is used in ERTS (HASHMAP_ESTIMATED_HEAP_SIZE(SIZE) == (SIZE*3 + (2*SIZE/5)*2)) but the result severely exaggerates the size of a map. Due to the difficulty anticipating the in-memory size of a map, the size is taken directly by using erts_debug:flat_size/1. The C macros above may change in the future (the macros above are from erts/emulator/beam/erl_map.h in Erlang 18.1).

BSD

Michael Truog (mjtruog [at] gmail (dot) com)

• Sverker Eriksson erts_debug:flat_size/1 info