Re: [MiNT] Where shall we go tomorrow?

[kellis <kellis@leading.net>]

On Sat, 11 Dec 1999, Martin-Eric Racine wrote:

> On 11.12.1999, kellis <kellis@leading.net> wrote:
> 
> > Now Martin E. Racine put something like: 
>                            ^^^^^^^^^^^^^^
> 
> Beware.  Those "something like" are what have made people say
> that I speak nonsense.  They don't pay attention nor quote me 
> then accuse me of saying things I never said.  Please don't.
Hence why I put "something like" but well. Besides people have gotten your
email as well. No rumours, just need to read what you wrote. Anyway.

> > However an FPU within the kernel, I am not too sure about
> > that one. How about DSP within the kernel might as well?
> 
> Notice I nowhere mention the kernel specificaly.  I was talking
> about using the -m68020, -m68881 or DSP, whenever it would make 
> packages more usable, in general.
Ah for packages?
well I think here it is up to the packager maintainer to decide what bin
he wants to release. 030, ok. ALthough, a "vi" package with 030 and FPU
only package is like hmmm whatever.. I don't think really clever. But
that'll work. I don't think it is that important. As those are compiler
optimization level. I was more thinking about OS enhencement for newer
hardware, newer TOS (in rom) based machine, and remove some of the fancy
stuff just to still support TOS 1.0. Which might clutter the MiNT src for
nothing, other than it is "cool" to know that MiNT could run on an 520 ST.
But no one would use it, as it is too slow, or magic would be their first
choice anyway.


> 
> I gave the example of SSH, one case of a package where -m68020
> and -m68881 flags increase performance enough to make it worth.
I think, to make it worth is up to the user to decide that.
CAB on a 520ST in magic with a 9600 bds modem in ST-LOW screen is not
worth it to browse the net. Yet some people don't mind it seems. Speed is
all relative. If it is only the "speed" part of it that you are
commenting, I think we are all in a pretty bad shape here. 
gcc on a 030 16mhz to compile can take days. When it takes less than 5 min
on an another platform. Yet we bare with it.

I am mostly concerned about current "work-around" as far as the 68000
compatibility goes.