threads
threads copied to clipboard
Semantics of bulk memory operations
The semantics of bulk memory operations on shared linear memory seems to be currently underspecified. If the memory model is mostly the same as JS's, I assume these should be unordered atomic.
AFAICT V8's current implementation is also unsafe in that it just calls out to C memmove and memcpy under the hood, which are non-atomic.
@conrad-watt should know best.