Re: Redundant type checking in window.c and w32menu.c

[Dmitry Antipov]

Jason Rumney wrote:

> dmantipov wrote:
> > Jason Rumney wrote
> > > Dmitry Antipov wrote:
> > >
> > >     
> > > > +#define Fcar(c) _FCAR (c)
> > > > +#define Fcar_safe(c) _FCAR_SAFE (c)
> > > > +#define Fcdr(c) _FCDR (c)
> > > > +#define Fcdr_safe(c) _FCDR_SAFE (c)
> > > >       
> > > How does lisp code then call these C macros?
> > >     
> > It will call original functions which are preserved in data.c. Lisp code calls 
> > them via pointers, so we definitely need to preserve an addressable versions.
> >   
>
>
> Is this optimisation really worth the confusion of having two versions
> of these functions in the code?

YMMV :-) - since these functions has the only arg, we shouldn't loose too much
on passing an arguments and performing the call. Probably this also depends
on compiler options passed and even from the compiler's version.

Another interesting thing to try (probably x86 only) is to use
__attribute__((regparm (1))) and see what happens...

Dmitry