On Wed, Apr 12, 2006 at 11:19:53AM +0100, Simon Marlow wrote:
I'll argue that point :) GHC makes run-time checks at safe points and implements preemptive concurrency. Cooperative scheduling is when the *programmer* has to insert the safe points.
the programmer of the standard libraries or low level FFI interfaces. not the end programmer. I would be highly surprised if anyone other than system or very low level library implementors ever actually needed to use an explicit 'yield'. certainly a whole lot less often than they have to add 'seq's and it is a lot more clear when they are needed :)
The safe points don't even have to be very often: in GHC the context switch check is made after every 4k of allocation.
indeed, which means GHC technically doesn't meet the preemptive requirements since a tight mathematical non-allocating loop can halt it. in order to do true preemption, you'd need to respond to SIGALRM or something like that, which can be quite tricky. John -- John Meacham - ⑆repetae.net⑆john⑈