On 2010-06-18, at 3:50 PM, Jecel Assumpcao Jr. wrote: > A single, huge associative memory would have the same problems as the > von Neuman architecture even though you have moved some of the smarts to > the memory side. Our brains are proof that you can have highly parallel > associative memories and my RNA proposal is an attempt to see what the > extreme case of that might be. Some hybrid scheme which replaced a > single 1 GB associative memory with 1000 such memories of 1 MB each > would be worth investigating. How would they collaborate?
More thoughts: IIRC, modern associative memories used in routers have ~256bit keys. This extra space for "pointers" allows you to do neat things like: 1) tagged memory - use part of the key to specify if the value holds a pointer (another key). If the hw can quickly scan the keys and values, you can implement a straightforward hw based garbage collector. 2) eliminating the MMU, radically simplifying dlls, linking, etc which are all ugly hacks to get associative lookup behavior 3) having each process use a, say, random 64bit prefix (generated by the OS/VM) for it's writes such that they would naturally be protected from other programs without explicit memory protection. A pipe could likewise be constructed and the keys for the pipe shared by the communicating processes. Security would become both a simple and natural function of the hw. _______________________________________________ fonc mailing list [email protected] http://vpri.org/mailman/listinfo/fonc
