[foreign] RFR Add sizeof method to Struct type

[Samuel Audet]

I saw that message, yes... So, it sounds like users will have to map
everything manually in Java? There won't be a way to map the C++ new and
delete operators, for example?

Samuel

2018年7月27日(金) 7:28 John Rose <john.r.rose at oracle.com>:

> On Jul 26, 2018, at 2:41 PM, Samuel Audet <samuel.audet at gmail.com> wrote:
>
>
> BTW, how memory gets allocated isn't limited to alignment. With CUDA…
>
>
> And there is also Intel XPoint memory, and lots of other kinds.  Assuming
> they
> are linearly addressed we can use pointers and layouts on all kinds of
> memory.
>
> We plan that the Scope API will handle any kind of closeable resource,
> which requires the kind of API you describe.  Right now it favors slices
> of vanilla native memory by having a methods for creating such slices.
> But those should be understood as convenience methods.  The correct
> API will be something like a static resource-factory object, which has
> kind-specific allocateInScope methods, and a handshake with the owner
> scope for managing lifetime.
>
> I described one aspect of this "future Scope" earlier this month:
>
> http://mail.openjdk.java.net/pipermail/panama-dev/2018-July/002317.html
>
> That message talks about "variables" and "objects"; we take it as given
> that many of these variables will be vanilla native memory, but some won't.
> The inter-scope handoff APIs are not designed yet, but will abstract across
> kinds of variables, allowing things like shared memory segments and exotic
> memory.  Scopes should also handle to external resources like open files,
> channels.
>
> We may well end up with Scope (perhaps renamed to Block) being
> a convenience class on top of a ResourceExtent API that manages
> all kinds of resources.  And any number of ResourceFactory
> implementations would hang their products on various ResourceExtents.
>
> — John
>
>