[Info-vax] The (now lost) future of Alpha.

[invalid]

On 2018-08-01, Tim Sneddon <tsneddon at panix.com> wrote:
> invalid <address at is.invalid> wrote:
>> On 2018-07-29, Simon Clubley <clubley at remove_me.eisner.decus.org-Earth.UFP> wrote:
>>> On 2018-07-29, invalid <address at is.invalid> wrote:
>>>> On 2018-07-24, Arne Vajh??j <arne at vajhoej.dk> wrote:
>>>>>
>>>>> Given that a compiler basically reads a text file and writes a file
>>>>> (text if it outputs assembler and binary if it outputs object code),
>>>>> then I find it very difficult to see why a compiler would have to be
>>>>> written in assembler.
>>>>
>>>> It was the best language for the job and in a lot of ways still is.
>>>>
>>>
>>> It may just be me, but I don't fancy the idea of writing a LL(1) parser
>>> (assuming the language is suitable for LL(1) parsing), an optimiser or
>>> a code generator in assembly language. :-)
>> 
>> Once you've done it, you mostly don't have to do it again.
>> 
>> The PL/I compiler had over 200 passes at one time. That's a lot scarier than
>> writing any parser in any language ;)
>
> What? 200 passes? As someone who is likely more knowledgable than most in
> this particular arena, just what are you talking about?  Please, site
> examples and references, I'm very interested.

Look on bitsavers for the MVS manuals, then PLM (Program Logic Manuals) then
find PL/I

>
>> 
>>>
>>>> What should they have used to write a FORTRAN compiler in 1957 or 1966?
>>>> A COBOL compiler in 1959? A PL/I compiler in 1964?
>>>>
>>>
>>> High level languages were a _lot_ simpler back in those days. :-)
>> 
>> Maybe, but the point was there was no other choice of implementation
>> language in those days. And 50+ years later  we're still using those
>> languages (except for IBM FORTRAN, which is sadly lost in time at F77+)
>> now. Which is why the compilers are still mostly assembler. Except maybe for
>> C/C++ which may be heading towards self-hosting.
>
> Okay, I haven't paid a lot of attention to this thread

and this is the key...

>, but what you seem to be asserting here is that compilers are still (as
> now, 2018) written in assembler.

IBM compilers for the mainframe were and are written in assembler or PL/X
which is a high-level assembler. We have some evidence the C/C++ compiler
has gone awry and broke from the 50 year tradition.

> If that is the case, I think you really need to reconsider your statement.

Rather, you need to either read the thread for context and know what I'm
talking about and what you're talking about. The z/OS environment, hardware
and software is not like anything else.. or I should say nothing else is
like it. We write everything in assembler for good reasons.