[RFC] Support of IEEE modules, first draft for comments
Tobias Burnus
burnus@net-b.de
Sun Nov 10 18:42:00 GMT 2013
FX wrote:
>> The patch currently looks very x86-64 centric; I think it would be useful to make it more general.
> I canÂ’t provide AIX or SYSV code because I donÂ’t have access to those targets, but I can help someone work on it.
Well, you could write a patch for those and ask the AIX and
Solaris/FreeBSD maintainers for review. That's how I did it for my last
patches and it shouldn't be difficult. (Except that Solaris and FreeBSD
often slighly differ in type names.)
>> I also wonder whether not more parts could be moved to a virtual module of the compiler itself. Currently, only ieee_selected_real_kind is handled in the compiler itself.
> ThatÂ’s desirable in principle, but most of the features will have to be delegated to library functions anyway: we donÂ’t have (and, for what I can say, donÂ’t want to have!) target-specific code in the front-end.
I disagree: We want to have target-specific code in the front end – and
we have target dependent code there. But I agree to that extend that we
don't want to hard-coded target code but rely on the target macros/hooks
for that.
Existing target-dependence: Available integer/floating point types, all
ISO_C_Binding parameters, HONOR_INFINITIES, HONOR_NANS and some more.
Tobias
More information about the Fortran
mailing list