On 13 января 2008 Pavel Pervov wrote: > On 1/13/08, Gregory Shimansky <[EMAIL PROTECTED]> wrote: > > On 13 января 2008 Pavel Pervov wrote: > > <SNIP> > > > > Then VM can unpack the structure if it can provide some performance > > > > > benefits. > > > > In this case all calls to GC API would have to use double indirection, as > > opposed to having a number of function pointers each filled up > > individually. > > So I support Xiao-Feng's original approach with initializing each > > function pointer like in > > > > gc_alloc = gc_interface(gc_module, GC_ALLOC); > > Just as I said - unpack on VM side.
Ah I didn't read what you wrote attentively. You are right, initializing a whole table of pointers is possible with such approach. -- Gregory
