On Wed, Feb 22, 2012 at 07:48:38PM -0500, Jonathan M Davis wrote: > On Wednesday, February 22, 2012 16:45:29 H. S. Teoh wrote: > > But still, some benchmarks do appear to be showing signs of a large > > performance hit on the GC when there happens to be many integers that > > look like valid pointers. > > As I understand it, this is mitigated considerably on 64-bit platforms > due to the large pointer size. [...]
True. You'd have to deliberately want to break the GC in order for coincidental integer values to cause a significant problem. But this problem could be a major issue if D was to become a significant player on handhelds, which, if I understand correctly, are still largely running 32-bit CPUs. It's also a major problem on 16-bit platforms, but the only use of those that I can conceive of are toy applications so it's probably not worth the consideration. :) T -- That's not a bug; that's a feature!
