On Mon, May 2, 2011 at 4:35 PM, Michael Matz <m...@suse.de> wrote:
> Hi,
>
> On Mon, 2 May 2011, Jan Hubicka wrote:
>
>> ! d_entry.base.from = t;
>> ! slot = htab_find_slot (cache->node_map, &d_entry, INSERT);
>> ! if (*slot == NULL)
>> {
>> /* Determine the next slot to use in the cache. */
>> if (insert_at_next_slot_p)
>> ix = VEC_length (tree, cache->nodes);
>> else
>> ix = *ix_p;
>> !
>> ! entry = (struct tree_int_map *)pool_alloc (cache->node_map_entries);
>> ! entry->base.from = t;
>> ! entry->to = ix;
>> ! *slot = entry;
>>
>> lto_streamer_cache_add_to_node_array (cache, ix, t);
>>
>> --- 348,367 ----
>> bool insert_at_next_slot_p)
>> {
>> void **slot;
>> unsigned ix;
>> bool existed_p;
>>
>> gcc_assert (t);
>>
>> ! slot = pointer_map_insert (cache->node_map, t);
>> ! if (!*slot)
>
> ix might legitimately be zero. Hence this transformation is not
> equivalent. You might want to enter ix+1 into the cache with the
> appropriate adjustment at read-out. Same for the other places.
Or not use index zero. Maybe better than also have to deal
with ix + 1 wrapping ...
Richard.
>
> Ciao,
> Michael.
>
