Re: [MiNT] GCC return register

[Paul Wratt <paul.wratt@gmail.com>]

2011/1/25 Vincent Rivière <vincent.riviere@freesbee.fr>:
> Miro Kropáček wrote:
>>
>> I said my opinion, the best fix is in my eyes completely forbid library
>> mixing.
>
> We have 2 solutions:
>
> - Allow library mixing, by patching GCC, so all binary interfaces will be
> identical and interoperable. This means that 68020-60 functions will return
> floats into d0/d1.
>
> - Forbid library mixing, best using Patrice's patch to make rules more hard.
> In that case having different interfaces for 68000 and 68020-60 is no more
> dangerous. But assembler sources must be aware of the differences.
>
>> No need to fix anything
>
> No, we have to fix something for 68020-60 (GCC or libs).
> The proof is that POV-Ray for 68020-60 currently does not work.
> Again, I have identified all the places requiring patching.
> The question is only to choose one solution or another.
>
> --
> Vincent Rivière
>
Maybe this will clarify a decision:

will a binary compiled using 68020-60 break on 020-040 that does not
have an FPU?

to me, from the people have made, d0 has more benefits than not, and
does not require the programmer to know library cavets.

With Patrice patch and other info here (on list) is it possible to
have solution #2 with a (simple) way to achieve no lib mixing?.

If returning in fp0 breaks on 020-040 that does not have an FPU then
that part of the answer is simple.

Paul