Re: [PATCH] Fix PRE iteration in PR40761

[Michael Matz <matz at suse dot de>]

Hi,

On Wed, 3 Mar 2010, Richard Guenther wrote:

> Bootstrap and regtest running on x86_64-unknown-linux-gnu (I wonder 
> which other users of postorder_compute rather want pre-order on the 
> reverse flowgraph ...).
> 
> If someone can suggest a more proper name for the function I'll move it 
> to cfganal.

pre_order_rcfg_compute?

I find the _compute suffix of all these functions mildly irritating, 
perhaps instead rename them all to get_xxx?  In which case it would be 
possible to include "cfg_" or "rcfg_" as a second prefix: 
get_cfg_post_order, get_rcfg_pre_order, and the like.

> + /* Compute a reverse post-order in *POST_ORDER.  If INCLUDE_ENTRY_EXIT is
> +    true, then then ENTRY_BLOCK and EXIT_BLOCK are included.  Returns
> +    the number of visited blocks.  */

The comment doesn't match now anymore.  And the function as now only works 
correctly if there aren't any reverse-unreachable blocks, that should 
probably be documented (or even gcc_asserted) or fixed.

Shouldn't this rather be factored with the current code in cfganal.c, 
where it then would be enough to have

workhorse (int *pre, int *post, bool on_rcfg)
...
> +       /* Look at the edge on the top of the stack.  */
> +       ei = stack[sp - 1];
> +       src = ei_edge (ei)->src;
> +       dest = ei_edge (ei)->dest;

  if (on_rcfg)
    src = ei_edge (ei)->dest, dest = ei_edge (ei)->src;
  else
    src = ei_edge (ei)->src, dest = ei_edge (ei)->dest;
  ...

and the rest of the code could be used as is?  (plus the obvious wrappers 
then).


Ciao,
Michael.