5 Dec
2018
5 Dec
'18
4:24 p.m.
Another DEC compiler that I forgot was the original C compiler for
Tru64 Unix on the Alpha. This was done at the DECwest facility in
Seattle (which originally had been set up by Dave Cutler). It was a
very strict implementation of the ANSI C89 standard--it had no
extensions such as K&R support. One customer called it the "Rush
Limbaugh of C compilers" because it was extremely conservative and you
couldn't argue with it.
-Paul W.
5 Dec
5 Dec
4:46 p.m.
Another curious compiler at DEC was VAX Ultrix Fortran. DEC had
gotten a lot of push-back from the research community, who wanted to
use VAX and Ultrix but considered the f77 compiler inadequate and
wanted to use VAX Fortran, which only ran on VAX/VMS. There was a
rush project to port VAX Fortran to Ultrix. It was decided that the
quickest way to get a quality compiler to market was to have the VAX
Fortran compiler continue to emit VMS-format object files, and to
modify the VAX/VMS linker to accept a.out object files and to emit
a.out images. Four of us worked on the linker port. Two of us from
the VAX/VMS languages team did the linker mods and two engineers from
the Ultrix group wrote code to translate VMS debug information to Unix
stabs. The resulting linker was called lk. The VAX Fortran RTL was
also ported, and since we had a way to produce Unix executables from
VMS object files, it meant we didn't have to rewrite the RTL, which
was mainly in BLISS but also had modules in Fortran, VAX assembler,
and Pascal.
-Paul W.
5:06 p.m.
On Wed, Dec 5, 2018 at 9:48 AM Paul Winalski <paul.winalski(a)gmail.com>
wrote:
Another curious compiler at DEC was VAX Ultrix
Fortran. DEC had
gotten a lot of push-back from the research community, who wanted to
use VAX and Ultrix but considered the f77 compiler inadequate and
wanted to use VAX Fortran, which only ran on VAX/VMS. There was a
rush project to port VAX Fortran to Ultrix. It was decided that the
quickest way to get a quality compiler to market was to have the VAX
Fortran compiler continue to emit VMS-format object files, and to
modify the VAX/VMS linker to accept a.out object files and to emit
a.out images. Four of us worked on the linker port. Two of us from
the VAX/VMS languages team did the linker mods and two engineers from
the Ultrix group wrote code to translate VMS debug information to Unix
stabs. The resulting linker was called lk. The VAX Fortran RTL was
also ported, and since we had a way to produce Unix executables from
VMS object files, it meant we didn't have to rewrite the RTL, which
was mainly in BLISS but also had modules in Fortran, VAX assembler,
and Pascal.
Nice work! We ran VMS on our MicroVAX rather than Ultrix or 4.[23]BSD
because of the FORTRAN compiler being so much better on VMS and our group
needing it for its hydrological simulations.
Warner
5:58 p.m.
I'll never forget Bill Munson's warnings at the 1980 USENIX conference.
Munson ran 'TIG' - Telephone Industries Group, in Merrimack, NH - who job
at DEC was helping their largest customer: AT&T. Us university types
were screaming at Bill and team "When is DEC going to 'support' UNIX?"
Bill
got up an cautioned, 'Be careful what you wish/ask us to do. If we do
support Unix, we will have to put Fortran, Cobol, and PL/1 on it -- which I
don't think you really want.'
I'm not sure which languages did eventually get supported and on which
versions of Ultrix. But once Paul's lk was released (and you can still
find it in /bin on the base Ultrix distributions), you did indeed see a
number of the languages move to Ultrix. I think for the Vax it was just
VAX/11C, Fortran and Pascal. I think Ultrix11 may have gotten Fortran, but
as I said; I don't remember. I do remember the TIG folks talking about a
PL/1 project and a proposal for Cobol and RP/G because some of the Wall
Street types wanted them, but I don't remember any of those getting
released (that said, I was also not watching things Vaxen by that time).
By the time I came back to Ultrix to do the MIPS 4000 stuff a few years
later, tech languages offerings were different and the GEM compilers had
come on the scene.
Clem
ᐧ
On Wed, Dec 5, 2018 at 11:48 AM Paul Winalski <paul.winalski(a)gmail.com>
wrote:
Another curious compiler at DEC was VAX Ultrix
Fortran. DEC had
gotten a lot of push-back from the research community, who wanted to
use VAX and Ultrix but considered the f77 compiler inadequate and
wanted to use VAX Fortran, which only ran on VAX/VMS. There was a
rush project to port VAX Fortran to Ultrix. It was decided that the
quickest way to get a quality compiler to market was to have the VAX
Fortran compiler continue to emit VMS-format object files, and to
modify the VAX/VMS linker to accept a.out object files and to emit
a.out images. Four of us worked on the linker port. Two of us from
the VAX/VMS languages team did the linker mods and two engineers from
the Ultrix group wrote code to translate VMS debug information to Unix
stabs. The resulting linker was called lk. The VAX Fortran RTL was
also ported, and since we had a way to produce Unix executables from
VMS object files, it meant we didn't have to rewrite the RTL, which
was mainly in BLISS but also had modules in Fortran, VAX assembler,
and Pascal.
-Paul W.
6:16 p.m.
On 12/5/18, Clem Cole <clemc(a)ccc.com> wrote:
I'll never forget Bill Munson's warnings at
the 1980 USENIX conference.
Munson ran 'TIG' - Telephone Industries Group, in Merrimack, NH - who job
at DEC was helping their largest customer: AT&T. Us university types
were screaming at Bill and team "When is DEC going to 'support'
UNIX?" Bill
got up an cautioned, 'Be careful what you wish/ask us to do. If we do
support Unix, we will have to put Fortran, Cobol, and PL/1 on it -- which I
don't think you really want.'
TIG's mission was to insure that DEC gear did what AT&T wanted. For
Unix this mainly was assisting in ports to new DEC hardware and
writing device drivers. Their mission was distinctly NOT to develop
anything DEC-specific. They were obsessively paranoid about
that--they called such things "vendor traps". Our attitude in the
VAX/VMS languages and software tools group was very different. We
were all about making VAX/VMS different, and better, in the
marketplace in order to attract customers to our platform.
When DEC officially decided to do their own, fully supported version
of Unix (to be branded Ultrix), things happened exactly as Bill Munson
predicted. The VAX/VMS languages & tools group started planning for
an orderly port of all of our products to Ultrix. We got immediate
and fierce push-back from the Ultrix engineering group (formerly TIG).
So we dropped the whole idea. Only to have to do a panic port of VAX
Fortran a few years later. Sigh. The wonder of culture clashes.
It took a long time for the "vendor trap" idea to die in DEC's Unix
group. Sun and others weren't as prissy when it came to putting their
own extensions into their Unix offerings, and the DEC Unix group
eventually found themselves being considered a trailing-edge offering.
-Paul W.
P.S. - I forgot VAX RPG as one of DEC's language products. I think it
was VCG-based, but I'm not certain of that.
6:26 p.m.
On Wed, 5 Dec 2018 at 13:01, Clem Cole <clemc(a)ccc.com> wrote:
I'm not sure which languages did eventually get
supported and on which
versions of Ultrix. But once Paul's lk was released (and you can still
find it in /bin on the base Ultrix distributions), you did indeed see a
number of the languages move to Ultrix. I think for the Vax it was just
VAX/11C, Fortran and Pascal. I think Ultrix11 may have gotten Fortran, but
as I said; I don't remember. I do remember the TIG folks talking about a
PL/1 project and a proposal for Cobol and RP/G because some of the Wall
Street types wanted them, but I don't remember any of those getting
released (that said, I was also not watching things Vaxen by that time).
By the time I came back to Ultrix to do the MIPS 4000 stuff a few years
later, tech languages offerings were different and the GEM compilers had
come on the scene.
http://bitsavers.trailing-edge.com/pdf/dec/vax/ultrix-32/4.0_Jun90/AA-MG63B…
seems to imply that as of Ultrix 4.0, there were COBOL and Ada compilers
for the VAX (Table 4-1). Elsewhere VAX LISP for Ultrix is mentioned, which
I had no idea existed. I hope that these compilers have been preserved
somewhere, as I imagine they sold in relatively small quantities.
-Henry
6:46 p.m.
On Wed, Dec 5, 2018 at 1:29 PM Henry Bent <henry.r.bent(a)gmail.com> wrote:
http://bitsavers.trailing-edge.com/pdf/dec/vax/ultrix-32/4.0_Jun90/AA-MG63B…
seems to imply that as of Ultrix 4.0, there were COBOL and Ada compilers
for the VAX (Table 4-1). Elsewhere VAX LISP for Ultrix is mentioned, which
I had no idea existed. I hope that these compilers have been preserved
somewhere, as I imagine they sold in relatively small quantities.
It all gets fuzzy by then - what was released and what actually existed.
Their was a Vax LISP from the Franz guys, but I think the DEC/CMU Common
LISP was VMS also was there. The 11 has been dropped from support an
the PMAX has replaced it but the compilers for PMAX are coming primarily
from MIPS. I do not believe DEC did a LISP themselves for the MIPS
machines. But, it turns out a number of the MIPS compilers would work on
Ultrix, although DEC had not released them directly. But some 3rd parties
did on their own, sometimes with DEC's blessing and sometimes not.
Plus the Alpha is being developed by then, so TLG is up to neck in the GEM
development cycle which would replace everything i the languages teams when
the dust settled (as Paul has mentioned it was an amazing piece of work at
the time and still somewhat lasts today - the VMS,Inc folks of course are
built an INTEL*64 code generator for it and supposedlty have VMS booting on
modern HW these days).
Clem
ᐧ
6:56 p.m.
Clem Cole wrote:
Henry Bent wrote:
Franz Lisp is kind of a Maclisp for VAX. It can run Macsyma.
There was also NIL, for VMS. Was it ever finished?
Elsewhere VAX LISP for Ultrix is mentioned, which
I had no idea
existed.
Their was a Vax LISP from the Franz guys, but I think the DEC/CMU
Common LISP was VMS also was there.
7:15 p.m.
On 12/5/18, Clem Cole <clemc(a)ccc.com> wrote:
Plus the Alpha is being developed by then, so TLG is up to neck in the GEM
development cycle which would replace everything i the languages teams when
the dust settled (as Paul has mentioned it was an amazing piece of work at
the time and still somewhat lasts today - the VMS,Inc folks of course are
built an INTEL*64 code generator for it and supposedlty have VMS booting on
modern HW these days).
One of the design goals for GEM was to avoid another fire drill like
the VAX Fortran Ultrix port. GEM had a set of modules called the GEM
Shell that isolated the rest of the compiler from OS
platform-dependent activities such as command line parsing, file I/O,
heap memory management, source locator management, and object file
generation. The GEM Shell presented a set of abstract APIs for these
activities to the rest of the compiler. This made support of new OS
platforms or object file formats relatively easy.
GEM already supported Itanium by the time Compaq sold off the Alpha
technology to Intel. VMS is mostly implemented in BLISS and VAX
assembler, and GEM-based Itanium-targeted compilers already existed
for both of these. VMS Software Inc. uses the GEM-based compilers for
support and ongoing new development on the Alpha and Itanium
platforms. For the Intel64 port of VMS, VSI is using Clang/LLVM as
their C/C++ compiler. For BLISS and VAX assembler, they are using the
GEM-based front ends, and a GEM-to-LLVM IL translator to use LLVM as
the optimizer/code generator.
-Paul W.
9:35 p.m.
As one of the OSF/1 T1.0 customers back when I remember well all these compilers during
the transition from Ultrix…
I was wondering if anyone remembers the HPF compiler and the “pangolin” parallel debugger
which I used around 1994 on an Alpha farm connected with FDDI.
I fondly recall our DEC contact for pangolin who was really helpful but I totally forgot
his name :( The HPF group in contact with us was also small but fixed loads of bugs we
reported.
Arrigo
7 Dec
7 Dec
7:03 p.m.
On 12/5/18, Arrigo Triulzi <arrigo(a)alchemistowl.org> wrote:
I was wondering if anyone remembers the HPF compiler and the “pangolin”
parallel debugger which I used around 1994 on an Alpha farm connected with
FDDI.
IIRC the DEC HPF compiler used the COMPASS Fortran 90 parser and front
end with GEM as the back end. That product lives on as the Intel
Fortran compiler, but with Intel's back end instead of GEM.
-Paul W.
1961
days inactive
1963
days old
10 comments
6 participants
participants (6)
-
Arrigo Triulzi -
Clem Cole -
Henry Bent -
Lars Brinkhoff -
Paul Winalski -
Warner Losh