On Fri, Mar 8, 2013 at 10:14 AM, Edward Z. Yang <ezyang@mit.edu> wrote:
Excerpts from Alexander Kjeldaas's message of Fri Mar 08 01:09:27 -0800 2013:
> The API has ccsDynamic, while the example uses newCostCentreStack.

Oops, fixed. (Added the newCostCentreStack helper function).

> Would it be possible to attach a listener to GC-events and dynamically
> adjust the listenCostCentreStack to trigger before major GCs?

Yes, but these listeners would only trigger after the GC, and have no
guarantee of triggering in a timely fashion (e.g. before the next GC).
Generally it's only safe to trigger "one-time" events.


Regarding "timely", can this be exploited by a Safe Haskell "tenant" to obstruct invocation of the listener to exceed resource limits?  Are there any guarantees that can be given? 

Here are some random thoughts:
During a GC, maybe the thunk attached to the cost centre stack could be put behind the listener on the same capability.

Or with the tickets system, maybe the listener could steal tickets from the thunk at every GC until it has executed, or something like that.

Or a thunk could be associated with an allocation budget, independent of the resident size, and when the allocation budget is exceeded, it will block on an MVar.  The listener/monitor has to run to increase the allocation budget.

Alexander
 
> Is this secure in the Safe Haskell sense?  Can this be used to create a
> secure computation wrt resources?  Could we listen to MUT seconds?

That's the idea!  But we'll need semantics before we can say for sure.
Actually, as far as time resources are concerned, the hope is to use
this in conjunction with http://hackage.haskell.org/trac/ghc/ticket/7606

Edward