On Thu, Mar 07, 2013 at 10:51:14PM +0000, Richard Sandiford wrote:
> void
> -save_optabs_if_changed (tree fndecl)
> +init_tree_optimization_optabs (tree optnode)
> {
> - /* ?? If this fails, we should temporarily restore the default
> - target first (set_cfun (NULL) ??), do the rest of this function,
> - and then restore it. */
> - gcc_assert (this_target_optabs == &default_target_optabs);
> + /* Quick exit if we have already computed optabs for this target. */
> + if (TREE_OPTIMIZATION_BASE_OPTABS (optnode) == this_target_optabs)
> + return;
>
> + /* Forget any previous information and set up for the current target. */
> + TREE_OPTIMIZATION_BASE_OPTABS (optnode) = this_target_optabs;
> struct target_optabs *tmp_optabs = (struct target_optabs *)
> - ggc_alloc_atomic (sizeof (struct target_optabs));
> - tree optnode = DECL_FUNCTION_SPECIFIC_OPTIMIZATION (fndecl);
> + TREE_OPTIMIZATION_OPTABS (optnode);
> + if (tmp_optabs)
> + memset (tmp_optabs, 0, sizeof (struct target_optabs));
> + else
> + tmp_optabs = (struct target_optabs *)
> + ggc_alloc_atomic (sizeof (struct target_optabs));
>
> /* Generate a new set of optabs into tmp_optabs. */
> init_all_optabs (tmp_optabs);
If TARGET_OPTIMIZATION_OPTABS is non-NULL upon entering the function,
then we call memset above, and then
...
/* If the optabs changed, record it. */
if (memcmp (tmp_optabs, this_target_optabs, sizeof (struct target_optabs)))
{
if (TREE_OPTIMIZATION_OPTABS (optnode))
ggc_free (TREE_OPTIMIZATION_OPTABS (optnode));
TREE_OPTIMIZATION_OPTABS (optnode) = (unsigned char *) tmp_optabs;
As tmp_optabs == TREE_OPTIMIZATION_OPTABS (optnode); in that case,
the ggc_free is wrong, TREE_OPTIMIZATION_OPTABS (optnode) will then point
to freed memory.
So I think the
if (TREE_OPTIMIZATION_OPTABS (optnode))
ggc_free (TREE_OPTIMIZATION_OPTABS (optnode));
has to be removed (of course the second ggc_free, if memcmp returned 0, is
desirable). Otherwise looks good.
Jakub
