On 2008-12-13 16:08, Antoine Pitrou wrote:
> Antoine Pitrou <pitrou@free.fr> added the comment:
> 
> Here is a new patch without any dispatch shortcut in ceval.c, just
> optimizations in unicodeobject.c and longobject.c. Net result on pybench:
> 
> Test                             minimum run-time        average  run-time
>                                  this    other   diff    this    other 
>  diff
> -------------------------------------------------------------------------------
>                  CompareFloats:   166ms   170ms   -2.3%   169ms   174ms
>   -2.8%
>          CompareFloatsIntegers:   230ms   231ms   -0.7%   233ms   231ms
>   +0.8%
>                CompareIntegers:   247ms   270ms   -8.7%   248ms   272ms
>   -9.0%
>         CompareInternedStrings:   196ms   254ms  -22.7%   197ms   255ms
>  -22.7%
>                   CompareLongs:   143ms   158ms   -9.0%   143ms   158ms
>   -9.3%
>                 CompareStrings:   156ms   168ms   -7.4%   157ms   169ms
>   -7.2%
> -------------------------------------------------------------------------------
> Totals:                          1139ms  1252ms   -9.1%  1148ms  1260ms
>   -8.9%
> 
> 
> The patch seems fairly uncontroversial to me, I'll commit it soon if
> there's no opposition.

Why have you removed the complete error handling section in
PyUnicode_RichCompare() ?

> Added file: http://bugs.python.org/file12341/cmps5.patch