Possible quick win in GC?

[thedeemon via Digitalmars-d]

On Monday, 29 September 2014 at 11:36:51 UTC, Abdulhaq wrote:

> So, (does anyone know) has this technique been discarded for D 
> or is it 'just' a matter of the resources to do it?

 From the literature on this topic I remember attempts for 
automatic region inference mostly failed: it lead to some small 
regions here and there that didn't affect anything much and then 
one huge region where most of the data landed, requiring a 
full-blown GC inside that big region. Making it work in D, where 
everything can be mutated by anything and any type system wall is 
a cast() away from breaking, seems virtually impossible.