gyro@cymbal.reasoning.COM (Scott Layson Burson) (05/04/91)
Okay, okay, I dug up the Hieb/Dybvig/Bruggeman paper (PLDI '90), and yes, they thought of the idea that I just said I hadn't seen elsewhere: of bumping a pointer to transfer frames from stack to heap without actually moving them. But they *still* didn't put all the pieces together in the same way, to eliminate copying entirely. Carl Bruggeman has just pointed out to me that my approach is perhaps only justified for programs where each continuation does a relatively small amount of work before invoking the next one (so that any copying on continuation invocation would represent a lot of overhead). As it happens, that is precisely the kind of program I had in mind (not that I have a working example handy). Anyhow, having gotten the feedback I was looking for, I'll shut up on this topic. Thank you all for your patience. -- Scott