On Thu, Jul 10, 2014 at 10:32 AM, Raoul Duke <[email protected]> wrote:

> How might/can/will Bit-C strive to prevent there being an inflection
> point where the developer has to say, "ok, i can't get the performance
> i need like this, now i have to start noodling over the weird details
> and ripping up my nice abstractions to eke out the performance i need.
> ... come to think of it, this is teaching me to just go back to my
> c/++ roots."?
>

Performance details like this come up at all kinds of levels, not just
primitive datatypes vs boxed objects. The most inspiring paper I've ever
read on this topic is the (original?) Aspect-Oriented Programming
<http://cseweb.ucsd.edu/~wgg/CSE218/aop-ecoop97.pdf> paper. Sadly, AspectJ
and followup work seemed to bear little resemblance to the promise
suggested by the original aspect weaver. Maybe I don't understand AspectJ;
maybe the original example was too narrow to be generalized; more likely
it's just a really hard problem.
_______________________________________________
bitc-dev mailing list
[email protected]
http://www.coyotos.org/mailman/listinfo/bitc-dev

Reply via email to