Checking the mark bit can be quite cheap, though, particularly if you store the mark bits in their own array - the next several bits will be in-cache. Still, the overall performance is likely to vary quite a bit for either scheme depending on the project's other quirks. Mark-sweep has the advantage that it doesn't need to move objects, but copying is easier with variable-size allocations, etc.
Another disadvantage with lazy sweeping is I don't see a clear way to make it generational. (Right?)
Another disadvantage with lazy sweeping is I don't see a clear way to make it generational. (Right?)