In TSO, there is a write buffer for each core.
Read on the same core can access the buffer, but other cores cannot.
In this case, the memory state is a function from location to value, M, plus a buffer for each thread:
TSO transitions:
Note that the buffer is propagated to main memory in the order it is fired.
Related
- cache coherence 每个线程的store buffer在cache之前。另外，TSO中的total store order，指的并不是操作实际到达内存的顺序，而是visibility order（到达coherent cache）。