Re: 'struct window' cleanup #2

[Dmitry Antipov <dmantipov@yandex.ru>]

On 06/25/2012 06:22 PM, John Wiegley wrote:

> Dmitry, have you measured what speed/memory advantage this change represents?
> I understand the theory of why it may be valuable, but what has convinced you
> that it's a hotspot?

1. The main bottleneck of Emacs' C code is mark_object; profiling says it may
consume 25-30% of CPU time for some workloads like long byte-compile runs.
So, lesser calls to mark_object is always desirable.

2. It doesn't make sense to use Lisp_Object for the integer values which are:
   a) invisible from Lisp;
   b) always contains an integer values (no nil or t with special treatment,
      for example);
   c) always converted to C integer values for the sake of comparison.

Dmitry