Once we improve the interconnects, today’s datacenters will be tomorrow’s shared memory NUMA supercomputers.
-
-
The main challenge to this isn't the quality of the interconnect, it's that you don't have temporal consistency in distributed systems.
-
The Google Spanner paper is an interesting look at how to address this problem when you have Google scale money.
- 6 more replies
New conversation -
-
-
We should also be optimising code for loyalty, for the Zeon Empire grows ever stronger in the vast cosmospic.twitter.com/66FzR9FilA
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
I’m excited for full support of virtual memory and page faults on GPUs in future.Perf critical code will still care,but most exploration not
-
Imo scheduling,initializing and copying is biggest entry and maintenance barrier to parallel computing-on GPUs,multi cores,datacenters.
End of conversation
New conversation -
-
-
HPs The Machine have some interesting ideas for better interconnects
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
To quote Knuth, optimisation is the root of all evil. Most games "programmers" are anal about shit they don't even know are a problem.
- 1 more reply
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.