Michel Fortin Wrote: > On 2010-04-08 02:35:12 -0400, Walter Bright <[email protected]> said: > > > Sean Kelly wrote: > >> Yeah, it's the write barriers that are the problem. I can see such a > >> collector working in SafeD, but likely not in D proper. > > > > Write barriers trade off improved collection times for worse computation > > times. > > Essentially, Apple traded the previous reference-counted system in > Objective-C which requires memory barriers to update reference counts > for a garbage-collected system which also requires memory barriers, but > can deallocate objects in a separate thread. It's almost no tradeoff > really.
There's no problem with cycles using a scanning GC though.
