Hello,
with the optimizations for array (Zsh 5.3.1-dev-0):

    typeset -a array
    array[100]=""
    array[20,29]=( 1 2 3 4 5 6 7 8 9 10 )

we can now have somewhat "native" buffers in zsh. Above will not do
anything with elements outside <20,30> range.

One use case are 2-dimensional arrays, with e.g. the library I sent to
zsh-users (ID: 21926). Created are e.g. 10 fixed rows of 100 elements.
Then it's possible to fill rows by storing to computed indexes. Or to
fill entire row by storing to <row_begin,row_end>. 

The same has been done with strings (Zsh 5.3), and string can be a
native buffer too.

That said, when I test with my screen-saver zsh-morpho
(https://github.com/psprint/zsh-morpho), which uses hash tables for its
pixel native buffer, arrays are still slower. It is 55.9 s vs. 27.3 s.
Callgrind reveals the cause:

59,930,247,427  arrlen_le [/usr/local/bin/zsh-5.3.1-dev-0_O2]
15,675,075,695  szone_free_definite_size
[/usr/lib/system/libsystem_malloc.dylib]
11,983,363,082  tiny_malloc_from_free_list
[/usr/lib/system/libsystem_malloc.dylib]
 8,037,180,402  tiny_free_list_add_ptr
 [/usr/lib/system/libsystem_malloc.dylib]
 7,181,545,835  _platform_memmove$VARIANT$Nehalem
 [/usr/lib/system/libsystem_platform.dylib]
(full data in attachement)

The buffer has size 34x158=5372 elements. Without the optimizations that
are the topic, time is 367 s.
-- 
  Sebastian Gniazdowski
  psprint3@fastmail.com