Best C compiler for Amiga?

[nicholas]

The Atari guys have ported GCC 6.2 to 68k.

http://d-bug.mooo.com/beyondbrown/post/gcc-6/

[Kernel]

SAS/C is freeware....you can get it over from the EAB file server (I think).

- note I just got Eclipse CDT working with gcc and Cygwin (only took me 3 days), now where's that cross compiler

Is this actually freeware now?

[kolla]

The Atari guys have ported GCC 6.2 to 68k.

Just a bunch of hackers

[HaukeVB]

Is this actually freeware now?

No, SAS/C isn't free nor is it open source... sadly...

[Kernel]

Well guess I try to get a copy through one of the sale sites then.  I used to own it too... long ago back in the early 90's but sold it with everything :/

Stooooopid me!

[wawrzon]

The Atari guys have ported GCC 6.2 to 68k.

http://d-bug.mooo.com/beyondbrown/post/gcc-6/

we have 6.1.0 on aros, still there are some issues with it sometimes especially on 68k, im trying to locate;). usually it is amiga specific stuff, like third party mui code, with all its specifics that gets accidenatlly optimized away.

[wawrzon]

You can try CodeBlock on Linux / Windows.

btw thanks for resurrecting this thread, probably a valuable info on using gdb with an aros/amiga/68k, from jason/ ezrec further above. probably something i was looking for all the time.

[nicholas]

Just a bunch of hackers

Yes, they'll never amount to anything and their code will be rubbish because they didn't get paid for doing it.

[nicholas]

we have 6.1.0 on aros, still there are some issues with it sometimes especially on 68k, im trying to locate;). usually it is amiga specific stuff, like third party mui code, with all its specifics that gets accidenatlly optimized away.

Might be worth getting in touch with the Atari guys and collaborating where possible?

PS: Have you tried setting up a Godbolt server on Linux to use the aros-68k-cross-compiler to help find the optimization bugs?

http://godbolt.org

[guest11527]

Yes, they'll never amount to anything and their code will be rubbish because they didn't get paid for doing it.

Actually, the 68K code generator of recent gcc versions *is* rubbish.

It's a clear demonstration of the open source problem: Instead of keeping the brilliantly working and well tested code generator of the 2.95 release active and try to get it working with a compatibility layer, the whole compiler was turned upside down. Certainly on good purpose, as to support more recent architectures, but that broke the entire lower layer which was, actually, no longer properly tested and checked.

Yes, the new code might be all polished, nicer, easier to maintain, supports C++ in its latest version, and C99 ... but just doesn't work brilliantly on old platforms, sorry.

[Thorham]

It's a clear demonstration of the open source problem: Instead of keeping the brilliantly working and well tested code generator of the 2.95 release active and try to get it working with a compatibility layer, the whole compiler was turned upside down.

That is NOT the problem, people are the problem. People are ALWAYS the problem. This is like blaming the goto statement for making a mess of things, while the programmer made a mess of things (it's ALWAYS the programmer's fault).

They didn't have to turn everything upside down, they choose to.

Yes, the new code might be all polished, nicer, easier to maintain, supports C++ in its latest version, and C99 ... but just doesn't work brilliantly on old platforms, sorry.

I'm surprised it still supports 68k at all.

[kolla]

Yes, the new code might be all polished, nicer, easier to maintain, supports C++ in its latest version, and C99 ... but just doesn't work brilliantly on old platforms, sorry.

You have data to back this up? Debian/m68k is using gcc6 and they have built around 10k of packages now, as well as the kernel, glibc etc. using newer gcc (currently gcc-6.2 I think).

[kolla]

I'm surprised it still supports 68k at all.

m68k as embedded platform is still a commercial thing. But most of all this has to do with the ColdFire CPUs, and how gcc etc treats ColdFire and m68k as the same "family" of architectures. Improvements done for CF are most often also done for m68k.

[nicholas]

Actually, the 68K code generator of recent gcc versions *is* rubbish.

It's a clear demonstration of the open source problem: Instead of keeping the brilliantly working and well tested code generator of the 2.95 release active and try to get it working with a compatibility layer, the whole compiler was turned upside down. Certainly on good purpose, as to support more recent architectures, but that broke the entire lower layer which was, actually, no longer properly tested and checked.

Yes, the new code might be all polished, nicer, easier to maintain, supports C++ in its latest version, and C99 ... but just doesn't work brilliantly on old platforms, sorry.

So you've heavily used GCC 6.2 on 68k to write C++14 code?

And even if what you say is true, why is this an "open source problem"? Does the latest MSVC++ compiler spit out 8086 real mode code that is as good as a 20yr older version? If not, why? " Closed source problem "? Nope, just technology moving on and getting left behind.