[jmm-dev] ECOOP, JVMLS

[Doug Lea]

On 07/19/2014 07:52 AM, Doug Lea wrote:
>
> Some of us on this list will be at ECOOP (http://ecoop14.it.uu.se/)
> the week of July 28 and others will be at JVMLS
> (http://openjdk.java.net/projects/mlvm/jvmlangsummit/).
>


Those of us at ECOOP informally discussed a few issues here and there
in groups of 2/3/4. The wednesday ad-hoc meeting was attended by
(about 15) people interested in hearing about JMM efforts but not
(yet) directly participating efforts.  I gather that Aleksey's talk
and follow-ups at JVMLS were also mostly attended by those possibly
affected by revisions.  Welcome to those of you who have since joined
this list.

Here are a few notes I jotted down.

* Julian Dolby and Mandana Varizi (IBM) plan to release MemSAT, the
tool used in their PLDI 2010 paper as part of the WALA toolkit
(http://wala.sourceforge.net/wiki/index.php/Main_Page).  Among other
possible uses, this should be helpful in comparing non-obvious
outcomes of JSR133 model vs whatever we arrive at. Details
forthcoming.

* Everyone seems to strongly favor the idea of replacing final-field
specs with those that amount to guaranteeing release fences in
constructors.  Still lots of details (some depending on core model) to
work out though.

* There is apparently some ongoing work updating .Net/C# MM-related
specs. The only (slightly) informative link I found on it is:
http://blog.alexrp.com/2014/03/30/dot-net-atomics-and-memory-model-semantics/

* While it is still far from a rigorous definition, several
people seem to agree that it is useful to think about OOTA as "not
observed in any non-speculative execution obeying the weakest (least
ordered) possible set of constraints". Along those lines, some
proposals might be seen as algorithms to permit some forms of
speculation while still conforming to this.

* Someone (other than Hans!) reminded me that something like the
reachabilityFence included in the never-shipped Fences API
(http://gee.cs.oswego.edu/dl/jsr166/dist/docs/java/util/concurrent/atomic/Fences.html)
needs to be part of deliverables, and will further require a few other
JLS touch-ups.

-Doug