[CakeML-dev] FFI calls

Hugo Férée H.Feree at kent.ac.uk
Tue Jun 27 09:26:22 UTC 2017


It would be great indeed! Although I have no urgent need for it. I think the immutable string would be enough and easier to use.



Le 27 juin 2017 09:18:30 GMT+01:00, "Johannes Åman Pohjola" <pohjola at chalmers.se> a écrit :
>When you put it that way, I realize that I'd prefer to have FFI calls 
>taking one immutable string and one byte array. This avoids the
>issue and works just as well if not better for my purposes.
>Does anyone (Hugo?) actually need it to be two byte arrays?
>On 2017-06-27 10:05, Magnus Myreen wrote:
>> Hi Johannes,
>> It would be great if you could work on this!
>> I suspect this change will mostly interact with what Hugo is doing at
>> the moment on the more realistic filesystem modelling for the FFI
>> interface (on the fileioffi branch).
>> What should the semantics be when the byte arrays alias?
>> There has also been talk about allowing the user to pass an
>> immutable string to the FFI interface, and then requiring that
>> the FFI call does not change the string. This would avoid copying.
>> Could you add this call-FFI-with-string feature at the same time?
>> Cheers,
>> Magnus
>> On 27 June 2017 at 09:59, Johannes Åman Pohjola <pohjola at chalmers.se>
>>> Hi!
>>> Word on the street is that there's been talk about changing FFI
>calls so
>>> that they take two byte arrays (representing, respectively,
>>> and data) instead of one. Is anyone working on this currently? If
>not I'd be
>>> happy to do it; it would improve my quality of life considerably
>were this
>>> change implemented.
>>> Cheers /Johannes
>>> _______________________________________________
>>> Developers mailing list
>>> Developers at cakeml.org
>>> https://lists.cakeml.org/listinfo/developers
>Developers mailing list
>Developers at cakeml.org

Envoyé de mon appareil Android avec Courriel K-9 Mail. Veuillez excuser ma brièveté.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cakeml.org/pipermail/developers/attachments/20170627/36888cfa/attachment.html>

More information about the Developers mailing list