"Model 2" prototype status

Simon Ochsenreither simon at ochsenreither.de
Tue Aug 4 19:09:42 UTC 2015


> Other languages and platforms have attempted to unify objects and
> primitives -- with varying degrees of failure and success. Paul
> Philips' talks circa 2013 -- at JVMLS and Scala PNW -- highlight the
> challenges of trying to unify references and primitives on the JVM while
> retaining any sort of compatibility. The poor performance of Scala
> streams over "primitives" (see Aggelos Biboudis' paper on this) proceeds
> mostly from this attempt to unify; the result is that they can't prevent
> the boxing, and performance suffers. .NET took the path of unifying
> structs and objects (but regrettably, allowed their structs to be
> mutable.) This has been partially successful but also less than
> perfect; we've tried to learn what we can from their experience.
 
I have seen this claim quite a few times already, and I think it's not a fitting
characterization of the issue.
 
I'll try to clarify a bit:
 
* The whole issue with unification and the Any/AnyRef/AnyVal hierarchy is that
we (in Scala) have to fight the JVM for every tiny bit of additional
correctness.
Sometimes, it's just a case of diminishing returns given the constraint that we
want to have a execution model where vanilla class files are executed on the
JVM, and not some "ScalaVM inside the JVM" approach.
If the JVM provides better support for these things, most of the issues we see
in Scala will just go away for good.
* The performance of Scala collection operations in the paper you mentioned is
unrelated to that.
Constructing a data structure after each operation is expensive (and a bad idea
from my POV), boxing is probably the least pressing issue. The paper [10]
referenced by Biboudis doesn't support his claim in my opinion, either. Of
course, getting rid of boxing is great, but no amount of boxing elimination can
turn an O(n) collection design into an O(1) collection design. (Which is pretty
much what ScalaBlitz does, last time it looked: It tries to eliminate the
construction of intermediate data structures, of course things get boxed less if
less work is done, but that's kind of obvious.)
I think the .NET devs have done a pretty good job overall. Apart from allowing
mutability, not supporting Void/Unit and completely messing up the semantics of
static inside of generic classes, things work quite well. They are sidestepping
the whole variance issue, but I wouldn't claim that that's not a valid solution
for them.


More information about the valhalla-dev mailing list