At 08:14 PM 8/12/2002 -0700, Sean O'Rourke wrote: >On Mon, 12 Aug 2002, Melvin Smith wrote: > > >4) Parrot_Coroutine's 'init' is not longer used and can go away, I guess > > >I could remove it in a future patch ... ok so that's not a question > > > > I wish this wouldn't go away. I think passing the constructor argument > > for any PMC is a good optimization. How to do it flexibly is another thing. > >It's not the constructor argument, I think. Coroutines used to have two >different code pointers, one for the first call, and one for their yield
Eek, sorry. Its embarrassing to forget my own code. I remember now and I did read your patch. I liked the "feel" of the yield op, but who cares since its machine generated anyway? >point. When I removed those ops awhile back, I also simplified coroutines >to only use one code pointer. The first time a coroutine's called, it >points to wherever that coro starts. Afterwards, it points to the last >yield point. This seems to give us the old semantics without the extra >pointer (not a big deal). You are correct. My first implementation was meant to be easy to understand and lead the way for more hackers to improve. Mission accomplished. :) -Melvin
