On 8/8/07, Richard Blewett <[EMAIL PROTECTED]> wrote: > > > unless you think we should special case the known primitives (there > is some code in place special casing situations like boxing Nullable<T>). > Then we come back to Fabian's point - assuming you have a shared boxed > int, > how long do you expect to keep this shared int around? If not forever > (which > is possible if you think about the behaviour of interned strings) then you > have just created a bunch of gen2 garbage that the GC will have to work > very > hard to collect.
I seem to recall that CPython keeps a predefined collection of small integer objects so that they don't have to be constructed from scratch. It might be effective for the JITter to similarly keep a permanent but small set of integers -- say, from -1 to 255 -- so that boxing an integer in this range is just a simple index operation. On the other hand, I don't think I'd bother implementing something like this without extensive profiling that suggests the performace is worth the increased complexity. -- Curt Hagenlocher [EMAIL PROTECTED] =================================== This list is hosted by DevelopMentor® http://www.develop.com View archives and manage your subscription(s) at http://discuss.develop.com
