[TUHS] fortran compiler, in assembly, for pdp-11

[Clem Cole]

below..

On Tue, Mar 7, 2017 at 1:13 PM, ron minnich <rminnich at gmail.com> wrote:

> I was not clear, evidently, although part of this is my memory's fault :-)
>
> There was a fortran compiler on v6, written in assembly. I was wondering
> who wrote it.
>
>
> On Tue, Mar 7, 2017 at 9:34 AM ron minnich <rminnich at gmail.com> wrote:
>
>> I spent a year or so working on this in 1977. I was wondering who wrote
>> it.
>>
>> Funny but: I once had a compile fail on Motorola's MPL compiler, which
>> was written in fortran. It had so many continued comment lines that the
>> 16-bit column number went negative, and I got a fairly obscure error.
>>
>> Anyone remember who wrote it?
>>
>

A one point the DEC RT11 compiler was moved to V6 and V7 (I want to say
around 1978 or 79 ish).   It was written in a mix of BLISS-11 and PDP-11
assembly, although the UNIX code was moved over as pure assembly. I do not
believe the BLISS-11 runtime was ever directly ported.    IIRC, the "port"
was done by the "Commercial Union Leasing Corp" guys folks in NYC.  The
assembly code for same was released to the UNIX community - I think on one
of the early USENIX tapes, could be the first one, two or three.

One of the things "CULC" guys did was they had to write a PDP-10 simulator
so they run the BLISS compiler which only ran as cross compiler from a 36
bit system.   I remember the USENIX talk about it at the time, somebody
asked the speaker if ITS would boot on the simulator - which got a lot of
chuckles.  As I remember the process, they ran the PDP-10 version of the
BLISS/11 compiler under enough of a TOPS-10 (I think) simulator that they
could feed it the sources to the DEC FORTRAN compiler.  IIRC a recompile on
their 11/45 took overnight or something like that.   I remember that they
said that the process was not interactive.

Like Ron, I remember seeing and using the results the process.  I do not
remember ever seeing the simulator or the input to the process that created
the assembly for the compiler.  I also do not remember how much of the
compiler sources they had from DEC.   They may only have had the runtime
source and a few files / "enough" of the upper layers to hack it into
working "good enough" for UNIX.   If you remember when you used it, the
switches were parsed RT-11 style because all that was in the code they did
not/could not replace.  My guess is that was probably what is was.   DEC
must published the runtime as "open source" and they picked it up and
started hacking.  But I don't remember more than that.

That said, it was "good enough" port/hack of the compiler and its runtime
to make it all work.   I probably still have it somewhere on 9-track.

As an interesting aside, I just saw the lead (Rich Groves) from the DEC
team for that compiler last Wednesday at a social function (he has been
quite ill and I'm pleased to see him out and about).    They DEC Fortran
folks really were an amazing crew and that legacy continues.   I fear not
enough people really know how far back their legacy goes.  In the old days
I used to kid Rich, asking him if there were more Fortran developers than
there were Fortran users.   But the truth is, 40 years into my career,
while I still don't want to program in it, as a language, it still very
much pays my salary.

Ron as you know from your days at the labs, Intel would not nearly be as
successful in HPC if not for its developer tools and those tools all have
the DEC compiler DNA ground up and injected into them.


Clem ​
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://minnie.tuhs.org/pipermail/tuhs/attachments/20170307/70d8b954/attachment.html>