[OpenIndiana-discuss] running sunpro binaries

[Udo Grabowski (IMK)]

On 09/01/2024 23:50, Goetz T. Fischer wrote:
> yes okay, compatibility reasons. but my question was: "does it still make sense to use sunpro
> performance wise?"
> 

For Fortran, a clear YES, it's comparable to INTEL ifort performance,
which doesn't exist for illumos, but for C/C++, I don't know
(you typically don't use C/C++ for performance critical numerical
projects in science ...). But even there, it already misses
Fortran 2003/2017/2023 standards, so having gfortran fixed for
basic language conformance would be preferable.

> On Tue, 9 Jan 2024 23:43:19 +0100, Udo Grabowski (IMK) wrote:
>> On 09/01/2024 22:37, Goetz T. Fischer wrote:
>>> On Tue, 9 Jan 2024 11:40:53 +0100, Udo Grabowski (IMK) wrote:
>>>> For C/C++, I would absolutely prefer gcc/g++
>>>
>>> why?
>>
>> As said. the old sunpro doesn't support modern standards,
>> and the newer ones that do are not binary compatible anymore;
>> and for C++, it causes issues with lousily written packages
>> to not compile without nontrivial patches, while gcc just
>> happily let them through, see:
>>
> <https://www.oracle.com/application-development/te>chnologies/developerstudio-cplusplus-faq.html#Coding10>
>>
>> The other culprit is the standards-required name-mangling
>> uniqueness for each compiler, which don't let you mix
>> C++ libraries from different brands to be called from one
>> program (which was dropped from later C++ standards),
>> which then requires you to have more than one C++ brand
>> of libraries at hand to link (this was one of the obstacles
>> to solve to get python scipy/numpy linked when using the Sun
>> F90 compiler for the LAPACK/BLAS libraries, which took a lot
>> of tinkering with the python toolchain on an old oi151a9).
>>
>> So if you don't have specific requirements forced on you
>> by already existing libraries using the sunpro stuff,
>> I would recommend to always go with the gcc/g++ tools.
>>
>> Didn't use clang/llvm, so can't say much about that._______________________________________________
>> openindiana-discuss mailing list
>> openindiana-discuss at openindiana.org
>> https://openindiana.org/mailman/listinfo/openindiana-discuss
> 
> _______________________________________________
> openindiana-discuss mailing list
> openindiana-discuss at openindiana.org
> https://openindiana.org/mailman/listinfo/openindiana-discuss