Skip compact_pool if all blocks are almost full

[nwellnhof]

If all blocks except the top block are almost full, compacting isn't needed. The attached patch makes compact_pool return early in this case.

Also, pad_pool_size currently includes pools that are almost full. This means that more memory is allocated than needed. It isn't a big problem because most of the time the memory will be used sooner or later, but it's cleaner to account for the almost full blocks here.

Plus a micro optimization in is_block_almost_full.

[chromatic]

On Friday 13 August 2010 at 15:31, Parrot  wrote:

>  If all blocks except the top block are almost full, compacting isn't
>  needed. The attached patch makes compact_pool return early in this case.
> 
>  Also, pad_pool_size currently includes pools that are almost full. This
>  means that more memory is allocated than needed. It isn't a big problem
>  because most of the time the memory will be used sooner or later, but it's
>  cleaner to account for the almost full blocks here.
> 
>  Plus a micro optimization in is_block_almost_full.

The patch seems to do good things for me with Rakudo.  The startup improvement 
is small, but I can measure its positive effects.

-- c 

[nwellnhof]

It seems that my patch skips compacting if there are only completely empty memory blocks and almost full blocks. In this case the completely empty blocks should be freed. Here is a fix on top of the first patch.

[chromatic]

Thanks, applied as r48566.