[J3] Problem with contiguous

[Keith Bierman]

On Tue, Dec 11, 2018 at 11:05 AM John Reid via J3 <j3 at mailman.j3-fortran.org>
wrote:

> Van,
>
> It sounds to me as if the problem is that your vendor is providing an
> inefficient allocator. Should you not be moaning to the vendor, rather
> than asking for a language extension?
>
>
> While a good point, since there's no magic pixie dust, I think Van has a
valuable use-case. That the internal version was performant and easily
readable is a good hint ;>  Due to the complexity of the code, I suspect
JPL has already discarded the notion of a macro processing step.

I think it would have been helpful if the thread-safety was mentioned in
the original description, that wasn't an obvious requirement ...

Communicating memory hints is hard; and memory layout and usage patterns
haven't gone away as critical to performance (if anything, GPUs, FPGAs,
caches, NUMA ... have made it harder).
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.j3-fortran.org/pipermail/j3/attachments/20181211/7b3a9488/attachment-0003.html>