Re: mikachu/badarrays (Re: Valgrind testing, ideas)

[Sebastian Gniazdowski <psprint@xxxxxxxxxxx>]

On 4 czerwca 2017 at 02:49:10, Bart Schaefer (schaefer@xxxxxxxxxxxxxxxx) wrote:
> On Jun 2, 3:40am, Sebastian Gniazdowski wrote:
> }
> } Is there code available somewhere?
>  
> This has been pushed to the zsh git server as schaefer/badarrays

Tested it with the screensaver that has screen-buffer (24x105 in the test; pushed the code to https://github.com/zdharma/hacking-private/tree/master/array_tests).

It is 2 second speed up, from 10.15s to 8.01s. Expected more, but the numbers apparently just sum up this way, even though arrlen_le prevails in callgrind run (below). For hashes the speed is 7.91s, so it seems that hashes are already top-speed containers.

Badarrays:

3,143,167,205  itype_end
2,132,125,016  _platform_memmove$VARIANT$Nehalem [/usr/lib/system/libsystem_platform.dylib]
1,269,228,237  stringsubst
1,196,046,124  stringsubst'2
1,112,113,319  ingetc
  935,633,068  strcpy [/usr/lib/system/libsystem_c.dylib]
  928,176,451  strtod_l [/usr/lib/system/libsystem_c.dylib]
  877,921,877  dquote_parse

Normal (HEAD):

13,503,068,671  arrlen_le
 3,144,808,355  itype_end
 2,133,150,428  _platform_memmove$VARIANT$Nehalem [/usr/lib/system/libsystem_platform.dylib]
 1,487,072,023  arrlen
 1,268,544,885  stringsubst
 1,196,046,124  stringsubst'2
 1,112,113,371  ingetc
   935,739,112  strcpy [/usr/lib/system/libsystem_c.dylib]
   927,736,913  strtod_l [/usr/lib/system/libsystem_c.dylib]
   877,921,877  dquote_parse

-- 
Sebastian Gniazdowski
psprint /at/ zdharma.org