Ian Lance Taylor wrote: > That means that the only correct way to share a memory buffer between > threads is to use a mutex, and it means that the mutex code can not be > written in standard conformant C. The mutex code must use asm > statements to insert memory barriers as required (on the x86 a lock > prefix serves as a memory barrier (except that some specific x86 > processors have bugs, but I digress...)). This comment makes me curious, since I see (and work on) plenty of code that shares data structures between threads using only a mutex for protection and no memory barrier instructions of any kind. This code is using standard POSIX threading constructs and seems to work fine on multi-CPU and multi-core systems, both x86 and other flavors. Are you referring to a specific sort of memory sharing between threads here, or are you saying that you think that _all_ objects shared between threads need to be protected in this way?
