Re: Do not compute alias sets for types that don't need them

[Michael Matz <matz at suse dot de>]

Hi,

On Fri, 22 May 2015, Jan Hubicka wrote:

> Index: tree-streamer-out.c
> ===================================================================
> --- tree-streamer-out.c	(revision 223508)
> +++ tree-streamer-out.c	(working copy)
> @@ -346,6 +346,7 @@ pack_ts_type_common_value_fields (struct
>       alias-set zero to this type.  */
>    bp_pack_var_len_int (bp, (TYPE_ALIAS_SET (expr) == 0
>  			    || (!in_lto_p
> +				&& type_with_alias_set_p (expr)
>  				&& get_alias_set (expr) == 0)) ? 0 : -1);

I find such interfaces very ugly.  IOW, when it's always (or often) 
necessary to call check_foo_p() before foo() can be called then the 
checking should be part of foo() (and it should then return a conservative 
value, i.e. alias set 0), and that requirement not be imposed on the 
callers of foo().  I.e. why can't whatever checks you do in 
type_with_alias_set_p be included in get_alias_set?

> +     front-end routine) and use it.
> +
> +     We may be called to produce MEM RTX for variable of incomplete type.
> +     This MEM RTX will only be used to produce address of a vairable, so
> +     we do not need to compute alias set.  */
> +  if (!DECL_P (t) || type_with_alias_set_p (TYPE_MAIN_VARIANT (TREE_TYPE (t))))
> +    attrs.alias = get_alias_set (t);

And if the checking needs to go down the main-variant chain then this 
should be done inside type_with_alias_set_p(), not in the caller, 
otherwise even the symmetry between arguments of type_with_alias_set_p(xy) 
and get_alias_set(xy) is destroyed (but see above for why I think 
type_with_alias_set_p shouldn't even exist).


Ciao,
Michael.