4 ms·
Can someone elaborate on the meaning of this for future programming paradigms? From what I understand, we need to get rid of shared memory when we have more th
by julius 16y ago
Can someone elaborate on the meaning of this for future programming paradigms?
From what I understand, we need to get rid of shared memory when we have more than 1000 cores. Is that correct? Should I stop using Threads and prefer Processes whenever possible, now?
- phaedrus 16y agoI think so. Actually the mantra "processes, not threads!" is something I've found to be a good way to solve problems in general. For instance, when I needed to run 3 simulated microcontrollers in a simulator that supported circuit plugins but only one microcontroller per simulation, this new way of thinking led me to realize: instead of trying to make the single simulator support 3 cpus, simply run 3 copies of the simulator process (and use tcp/ip to link them, via custom circuit plugins). I think, that once we embrace separate processes for concurrency instead of threads, we'll begin to see how abandoning the shared memory model can actually make some things simpler.