Distinguish between pre-write and post-write barriers

[wks]

In #1355, we are adding a new Barrier implementation SATBBarrier. Unlike the ObjectBarrier, SATBBarrier is a pre-write barrier. Currently there is nothing in the types to distinguish between a pre-write barrier (e.g. SATBBarrier), a post-write barrier, or a write barrier that doesn't really care about whether it is called before or after the field write (e.g. ObjectBarrier).

Is this a problem?

One thing is clear that we need a way to tell the user where to call them, before or after the actual field write.

Related comment: #1355 (comment)

[qinsoon]

One thing is clear that we need a way to tell the user where to call them, before or after the actual field write.

In the current write barrier API, we don't need to tell the users if pre or post methods are implemented/empty. The users are supposed to call both pre and post write barrier. If they implement a inlined fast-path, they mimic what MMTk core does. We also don't need to tell them which is implemented/empty programmatically.

We can introduce a new API to replace the current write barrier API. See your issues: #1247 and #1038.