Re: RFA: Get rid of tree annotation

[Michael Matz <matz at suse dot de>]

Hi,

On Mon, 23 Nov 2009, Richard Guenther wrote:

> +#define DECL_VAR_ANN_PTR(NODE) \
> +  (TREE_CODE (NODE) == VAR_DECL ? &(NODE)->var_decl.ann \
> +   : TREE_CODE (NODE) == PARM_DECL ? &(NODE)->parm_decl.ann \
> +   : TREE_CODE (NODE) == RESULT_DECL ? &(NODE)->result_decl.ann \
> +   : NULL)
> 
> if you'd have put the ann field into tree_decl_with_rtl you could
> have avoided the above.  I doubt it would have made a big
> difference - in fact only LABEL_DECLs and CONST_DECLs
> would have an unused field that way.

And tree_decl_with_vis, hence tree_decl_non_common, hence FUNCTION_DECL, 
TYPE_DECL (!), and template and namespaces.

> So, any special reason you didn't go that way?

I wanted to go all the way in saving memory, not doing something fragile 
again where it's not clear if this flag is used in some context or isn't.


Ciao,
Michael.