1980s/1990s 68k C cross (and not so cross) compilers

[Brad Parker]

On 11/21/15 2:38 PM, Al Kossow wrote:
>
>
> On 11/21/15 10:44 AM, Paul Koning wrote:
>
>>> Arg, totally forgot to include the HP 64000 and Tek 8560 development 
>>> systems though I'm
>>> blanking right now on if they did their own or sold third-party C 
>>> compilers.
>>
>> Third party, I believe.  I used one of those for a 68040 (developing 
>> the DECbridge 900).  I think the compiler was Green Hills.  GCC was 
>> around, I think, but that isn't the one we used as far as I remember.
>>
>>     paul
>>
>
> It would have been impossible to use GCC on the 8560, it was a V7 
> PDP-11 Unix. The 64000 processor is pretty much the same as the HP 9845.
>
In 1983 there was Alcyon on the 8560 :-)  (woa.  that was a long time ago)
It only worked because they #ifdef'd out the floating point support.  
Split I&D on a pdp-11.  Not much space.

I'm not aware of any native TEK C compiler for 68k on the 8560.

In 1987 gcc would compile to 68k quite well.  Before that I seem to 
recall that there was also a C compiler from Standford, from sumex (wow 
- do I still have those brain cells?).  Remember sumex-aim ? SumMacC. 
Anyway, I think the Kinetics fastpath was compiled with that and I could 
swear I was using it as a C compiler on a vax-11/750 running mt. xinu in 
mid 80's.  Find someone from pixar - they were using it to compile 
Macintosh code.  I don't know the lineage of that compiler, but I think 
it was a port of something older.

There was also another C compiler we ran on the vax but I doubt I can 
remember who wrote it (was it Megamax C?).  It was a top down recursive 
decent compiler with only a peephole optimizer, but it was "good 
enough".  I think I shipped a product or two with its output. It had a 
native mac version but somehow we got them to port it to the vax since 
it was written in C (and OMG, all of the token values were hex values in 
the code - like 0x45 - no use of the preprocessor at all.  made my hair 
catch fire.  still, it worked.  just don't try and edit it.)

Oh, and the sun-2 showed up around 1984.  That had it's own C compiler, 
naturally, but (oddly), I never used it as a cross compiler.  I was too 
busy trying to keep the lisa profile disk going :-)  Around that time 
the lisa was macintosh development environment.  But the pain keeps me 
from remembering it well.   The text editor was pretty.  That's about 
all I can say.  slow as molasses.

No one mentioned Apollo.  Ok, their pascal compiler probably had more 
strength but I know there was a C compiler in there as well, certainly 
in the 1983-1985 time frame.  and who could forget Bill Poduska?

There was also that wacky company in cambridge which made a hybrid 
c-compiler/c-interpreter.  It worked really well but it wasn't quite 
right.  I loved the idea but never bought one.  That was around 1988, 
1989?  I always hoped to find the source code for that.

I think gcc was the standard for 68k from 1987 on.  Yes, greenhills, but 
it's not clear it produced better code and it was really expensive.  Ask 
anyone at cisco or wellfleet what they used.  And trust me, they were 
worried about code size and code speed.  lol.  I remember trying to 
route at "wirespeed" using 10baseT.  Makes me laugh now.

-brad