[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [MiNT] Still don't get it - difference between 68020 and 68020-60



On Sat, 2010-06-19 at 13:49 +0200, Vincent Rivière wrote:
> Patrice Mandin wrote:
> > That's why I keep a gcc patch around to be able to build 68020 code
> > without FPU, mainly to target standard Falcon without FPU.
> >
> > I understand why Vincent would not put it in his gcc patch, as it would
> > mean 8 multilib configurations (68000,68020,68020-60,m5475, each
> > with/without mshort) when building gcc.
> 
> My very own opinion:
> 
> When I started working on compiler stuff, my only concern was about having 
> an up to date C++ compiler to produce optimal code for ST. And I must thank 
> you once again because if you hadn't put your GCC 3.x patches on the web I 
> would never have started working on that.
> 
> So the only interesting target for me was 68000 with mshort for optimal 
> performance (mlong is required for UNIX stuff but I don't care). Since the 
> MiNTLib is the only available libc for TOS, I still have to use it along 
> with mlong, but I really would like to get rid of it and use another 
> TOS-only mshort library one day.
> 
> Currently the things have changed a bit, since I have to make tests with a 
> ColdFire evaluation board and I will have a FireBee when available. So I 
> want to produce optimal code for ColdFire, namely -mcpu=5475 -mlong.
> 
> That's all for me.
> 
> But maintaining GCC up to date is not a trivial task. That work can be 
> useful for Falcon users and MiNT people. And I'm proud to see that work is 
> starting to become the basis for new developments areas like SpareMiNT, 
> Gentoo, RPM, ACP, EmuTOS... Thus I'm ready to work a little more if it can 
> help the whole community.
> 
> I have been away from any Atari business for 10 years. I have never used or 
> even seen a Falcon. So I don't care about any 68030 targets. I don't know 
> how many Falcon have an FPU.
> 
> So, about the multilib choices, I really don't know myself what is useful or 
> not for the community. You have to decide together.
> 
> Recompiling all the GCC stuff takes a lot of time, especially on Cygwin. It 
> was a lot worse in the past, when the computers were less powerful and the 
> patches had compilation problems. So adding any multilib is not a problem 
> for me, it will not be a lot worse.
> 
> The only important thing is that we have to decide about the standard 
> multilib set for MiNT and use it for all libraries.

Agreed.

Realistically we have....

68000 (Stock ST)
68000+68881 (Mega STE with FPU)
68020 (Stock ST with accelerator)
68020+68881 (Stock ST with accelerator with FPU)
68030 (Stock Falcon)
68030+68882 (Stock TT)
68040 (Afterburner / Hades 040)
68060 (CT60 / Hades 060)

I think we should have them in multilib, at least in the GCC compiler.

Alan.