Re:Pessimism about parallelism
by zzzcpan on Wednesday December 26, 2018 @11:26AM
(#18762900)
That's stretching it quite a bit. You can implement shared memory multithreading on top of actor model and have all of the same issues. If you don't do that though, you won't have those issues. Same way that you can implement memory unsafe primitives on top of memory safe ones.
[ Parent
| on HN ]
I have seen deadlocks and race conditions in non shared memory concurrent systems many times. They are no way exclusive of shared memory systems or even more common. They have nothing to do with shared memory and everything to do with message ordering.
It doesn't have to be shared memory specifically, but it has to be concurrent use of something shared even if that something is completely virtual and only exists in your mind. Which then forces you to synchronize access to that something and deal with all those problems. It's not about message ordering though, not unless synchronization is involved.