Date: 2001-08-13 04:10:28
Jens Maurer <Jens.Maurer_at_[hidden]> wrote...
> Any "large" Unix machine nowadays is NUMA (non-uniform memory
> architecture), i.e. there's "fast" memory (close to the CPU
> where the thread is currently executing) and there's "slow"
> memory (close to some other CPU). This applies to both stack
> and heap.
Do these machines have language extensions that show how we need
to change the abstract machine, or is it all handled transparently
by the OS (as with virtual memory) or hardware (as with processor
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk