Yup, that was Steve Ward's folks in the MIT/RTS group - it was the NU computer work.  John Siber did most of the compiler work (funny, Steve Johnson and I were talking about some of that work last night at the UNIX50 party last night).  tjt wrote the 68K assembler ward's folks used.  I don't remember where the Z8000 assembler came, but I'm petty sure that the Intel assembler and some of the tools other John had brought back from his summers in MH.

I think (but don't know for sure) the Intel 8086 assembler was done at AT&T first.  IIRC it may have come out of Dale's group in Columbus.   I do know Dale's group had done a Z80 C Compiler using the Ritchie Compiler at some point in 1978 timeframe (and at one time I had, but can not seem to find it, in my archives).

When Intel released the 386, I believe the AT&T 8086 assembler was updated for the new 32 instructions; although who did that/where was done, I'm not sure.

Steve is probably the best source for most of this as he managed the team in Summit doing the different AT&T front and back ends when they tried to centralize the compiler work for UNIX.

On Thu, Jul 11, 2019 at 12:48 PM Warner Losh <imp@bsdimp.com> wrote:


On Thu, Jul 11, 2019 at 10:31 AM Clem cole <clemc@ccc.com> wrote:
By the time of 4.2 the switch from the  Ritchie and Johnson compilers at UCB had begun.  Remember the primary output of Rms at that point was emacs and gcc.    

CSRG wanted the different backends for C.   ThAts it.  Besides the vax, Rms had done 68000 and 386 back ends then.  

With the original system V, all of AT&T, Intel and IBM paid Interactive Systems Corp (aka ISC) to port the System V/Vax code to a 386 ps/2 and an Intel reference system that used an ISA bus.  This would be eventually released in source at the 386 port from AT&T.   As part of the contract summit supplied the compiler

I know the AT&T assembler with it’s backwards syntax from Intel was done before rms did his.  He was compatible with the summit assembler.  I don’t remember who’s 386 backend came out first.  I think is was the summit compiler but you needed a system v license which UCB did not have. 

There's also a fair amount of work at MIT to do Intel code generation for 8086 (small mode) that was extended by Queens College London (I think, I gotta grab the tapes again) to do large mode. I've run into this looking for a compiler for the Venix source restoration project I've been tilting at. I found those based on a cryptic comment I found somewhere online about the tech behind Venix that wasn't from AT&T. I don't know if ISC started with them as a base or not, nor really how the MIT compilers came about, but they claim to be PCC based somehow. Don't know if this helps you on your quest... BTW, I found these when I found the latest pcc-restoration sources didn't have a working i86 backend anymore (there was once one for Minux, but when I built it I couldn't get it to generate sensible code at all).

Warner
 
Clem

Sent from my PDP-7 Running UNIX V0 expect things to be almost but not quite. 

On Jul 11, 2019, at 8:50 AM, Jason Stevens <jsteve@superglobalmegacorp.com> wrote:

That would make sense.   I was able to find some info on PCC2 here

I'm guessing along with the adoption of emacs the csrg must have been further gnu synergy...  Or maybe PCC2 just wasn't available outside of the labs?

Or maybe by '88 gcc was already usurping many of the c compilers of the era.





On Thu, Jul 11, 2019 at 11:37 PM +0800, "Clem cole" <clemc@ccc.com> wrote:

I believe the pcc/386 came out of Steve Johnson team at Summit with the PCC2 work.  

Sent from my PDP-7 Running UNIX V0 expect things to be almost but not quite. 

On Jul 11, 2019, at 7:53 AM, Jason Stevens <jsteve@superglobalmegacorp.com> wrote:

Does anyone know where the 386 port from PCC came from?

 

While trying to build a Tahoe userland for the i386, it seems that everything was built with GCC…

Was there a PCC for the i386 around ’88-90?  It seems after the rapid demise of the Tahoe/Harris

HCX-9 that the non Vax/HCX-9 platforms had moved to GCC?

 

Also anyone know any good test software for LIBC?  I’ve been tracing through some

strange issues rebuilding LIBC from Tahoe, where I had to include some bits from

Reno to get diropen to actually work.  I would imagine there ought to have been some

platform exercise code to make sure things were actually working instead of say

building as much as you can, and playing rogue for a few hours to make sure

its stable enough.