- * these are _the_ atomic operations, currently cast atop CompareAndSwap,
- * which is implemented in assembler. if we are worried about the cost of
- * this layering (we shouldn't be), then all this stuff could be
- * implemented in assembler, as it is in MacOS8/9
- * (derived from SuperMario/NativeLibs/IO/DriverServices/Synchronization.s,
- * which I wrote for NuKernel in a previous life with a different last name...)
- *
- * native Boolean CompareAndSwap(UInt32 oldValue, UInt32 newValue, UInt32 * oldValuePtr);
+ * These are _the_ atomic operations, now implemented via compiler built-ins.
+ * It is expected that this C implementation is a candidate for Link-Time-
+ * Optimization inlining, whereas the assembler implementations they replace
+ * were not.