Messages in this thread | | | Date | Thu, 12 Jul 2018 16:43:46 -0400 (EDT) | From | Alan Stern <> | Subject | Re: [PATCH v2] tools/memory-model: Add extra ordering for locks and remove it for ordinary release/acquire |
| |
On Thu, 12 Jul 2018, Andrea Parri wrote:
> > It seems reasonable to ask people to learn that locks have stronger > > ordering guarantees than RMW atomics do. Maybe not the greatest > > situation in the world, but one I think we could live with. > > Yeah, this was one of my main objections.
Does this mean you don't think you could live with it?
> > > Hence my proposal to strenghten rmw-acquire, because that is the basic > > > primitive used to implement lock. > > > > That was essentially what the v2 patch did. (And my reasoning was > > basically the same as what you have just outlined. There was one > > additional element: smp_store_release() is already strong enough for > > TSO; the acquire is what needs to be stronger in the memory model.) > > Mmh? see my comments to v2 (and your reply, in part., the part "At > least, it's not a valid general-purpose implementation".). > > > > > Another, and I like this proposal least, is to introduce a new barrier > > > to make this all work. > > > > This apparently boils down to two questions: > > > > Should spin_lock/spin_unlock be RCsc? > > > > Should rmw-acquire be strong enough so that smp_store_release + > > rmw-acquire is RCtso? > > > > If both answers are No, we end up with the v3 patch. If the first > > answer is No and the second is Yes, we end up with the v2 patch. The > > problem is that different people seem to want differing answers. > > Again, maybe you're confonding v2 with v1?
Oops, yes, I was. v1 was the version that made RMW updates be RCtso. v2 and v3 affected only locking, the difference being that v2 used unlock-rf-lock-po and v3 used po-unlock-rf-lock-po.
Alan
| |