I disagree with this assertion: "we will need to embrace a style of programming without any synchronization whatsoever."
Rather, synchronization will just shift from the physical domain (e.g. compare-and-swap, data races on the network or bus) to the logical domain (e.g. temporal semantics, consistency through commutativity and logical monotonicity). This essentially means trading memory for parallelism and performance. Berkeley's Order of Magnitude project (BOOM) and my own work on RDP are examples of effectively supporting logical synchronization. But there are a lot of precedents, e.g. regarding various `synchronous programming` models. * http://lambda-the-ultimate.org/node/4162 * http://awelonblue.wordpress.com/2011/10/06/vat-model-for-rdp/ I also expect we'll shift latency into the logical domain. Regards, Dave On Mon, Jan 16, 2012 at 1:25 PM, Erik Terpstra <erterps...@gmail.com> wrote: > Not entirely sure if this is off topic or not, but here it goes: > > > http://my-inner-voice.blogspot.com/2012/01/many-core-processors-everything-you.html > > _______________________________________________ > fonc mailing list > fonc@vpri.org > http://vpri.org/mailman/listinfo/fonc >
_______________________________________________ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc