Re: [PATCH 4/7] Class-fy partial_die_info

[Yao Qi <qiyaoltc@gmail.com>]

Simon Marchi <simark@simark.ca> writes:

> From what I understand, the only reason to have that private constructor is
> to construct a temporary partial_die_info object used to search in the htab,
> is that right?  If so, converting that htab_t to an std::unordered_map would

Right.

>
> -  /* Hash table holding all the loaded partial DIEs
> -     with partial_die->offset.SECT_OFF as hash.  */
> -  htab_t partial_dies = nullptr;
> +  /* Hash map holding all the loaded partial DIEs
> +     with their section offset as the key.  */
> +  std::unordered_map<sect_offset, partial_die_info *> partial_dies;
>

This doesn't compile with my g++ 4.9, as library doesn't provide
std::hash<T> specialization for enumeration types.  It is available
since C++ 14.  http://en.cppreference.com/w/cpp/utility/hash

I can change it to

std::unordered_map<std::underlying_type<sect_offset>::type,
		     partial_die_info *> partial_dies;

to fix the compiler errors.

> @@ -18259,14 +18227,7 @@ load_partial_dies (const struct die_reader_specs *reader,
>    if (cu->per_cu->load_all_dies)
>      load_all = 1;
>
> -  cu->partial_dies
> -    = htab_create_alloc_ex (cu->header.length / 12,
> -			    partial_die_hash,
> -			    partial_die_eq,
> -			    NULL,
> -			    &cu->comp_unit_obstack,
> -			    hashtab_obstack_allocate,
> -			    dummy_obstack_deallocate);
> +  cu->partial_dies.clear ();

This changes the behavior, without this patch, cu->partial_dies's
elements are allocated on cu->comp_unit_obstack.  After this patch,
elements are allocated on heap.  I don't know how does it affect
performance.

-- 
Yao (齐尧)