Re: [patch] Lno branch merge -- scalar evolutions analyzer

[Zdenek Dvorak <rakdver at atrey dot karlin dot mff dot cuni dot cz>]

Hello,

> On Tue, Jul 06, 2004 at 03:50:02PM +0200, Zdenek Dvorak wrote:
> > 
> > from the similar reasons, it might be better to avoid also the
> > loop_from_num ... loop_exit_edge accessors (although it of course is
> > not big deal either, it is a bit confusing to have both direct accesses
> > to the structure and these wrappers in the same code).
> > 
> 
> I prefer to read clear code instead of pointer accesses.  I get more
> information from a function call than from a pointer access.  As an
> example:
> 
> *************** build_classic_dist_vector (struct data_d
> *** 1456,1462 ****
>   	&& init_v[lca_nb] == 0)
>         dist_v[lca_nb] = 1;
>       
> !     lca = outer_loop (lca);
>       
>       if (lca)
>         {
> --- 1448,1454 ----
>   	&& init_v[lca_nb] == 0)
>         dist_v[lca_nb] = 1;
>       
> !     lca = lca->outer;
>       
>       if (lca)
>         {
> 
> I think it is worth to replace the pointer accesses with clear
> function (or macro) names.

for me it gives *less* information (I don't know what other magic the
accessor function might do without looking it up).  But if the general
preference is for the accessor functions, I don't mind.

Zdenek