Re: Gentler Decimal Floating-Point


In article <eul33f$p4v$1$8300dec7@xxxxxxxxxxxxxxxx>,
"Mike Cowlishaw" <mfc@xxxxxxxxxx> writes:
|> >> PL/I's 'float decimal' (always allowed to be in lower case :-)) was
|> >
|> > I am pretty sure that PL/I(F), like all OS/360 compilers, requires
|> > upper case. I do remember putting lower case comments in, though.
|>
|> Yes, you're correct .. I meant today it's always allowed. Lowercase was
|> introduced with the Optimiser & Checker PL/I compilers. They came in in the
|> early/mid 1970s, so I doubt the F compiler is used much nowadays.

I doubt it VERY much! While it implemented quite a different language,
and so might linger on, support was dropped as soon as the Checkout
and (later) Optimizer stabilised, and it was not binary compatible
with those. By 1990, most people in Santa Teresa didn't even know that
it had ever existed!

|> No, this was around the time of the Checker and Optimiser, 1971. I was a
|> pre-uni student at the time, working on testing the compilers with a 'monkey
|> program' which generated syntactically and semantically valid 'random' PL/I
|> programs. But I remember the arguments about implementing decimal float
|> using binary (actually hex) floating-point, because I was designing my
|> second language at the time (I still have the 2500 cards that are the
|> compiler, which was written in PL/I), and I wanted to use a decimal numeric
|> system as it was intended for teaching programming at my old school...

Interesting. I can remember the debate from the customer side, where
we assumed that it had been added to the language as a bright idea by
someone with more seniority than numerical experience. It was very
clear that all of the requirements were much better met by the fixed
decimal types - let's exclude PL/I's problems with them, as they are
an artifact of PL/I not the arithmetic.

To be blunt, nobody wanted it. The commercial programmers wanted
decimal fixed-point, and numerical ones either wanted floating-point
in any sane base (i.e. any of 2-c.16) or binary floating-point. Of
course, both of the latter wanted rounding, but that is another story :-)


Regards,
Nick Maclaren.
.

Relevant Pages

  • Re: Problem with a linked list
    ... And vendors who make C compilers generally make C and C++ combined products. ... On the rare occasions they do change the intermediate language, they will change all the front ends to use it so they only have one optimiser and code generator to maintain. ... This is why basically no vendor is rushing to support the "C99" standard. ...
    (comp.lang.c)
  • Re: Java 7 features
    ... application domain that need a log function or floating-point ... way to do that without a log function. ... that one of my first assignments was getting the COBOL D and FORTRAN D ... compilers of DOS/360 to play nicely together, ...
    (comp.lang.java.programmer)
  • Re: Paper on PL/I
    ... >> C also has the advantage of being married to the Unix architecture, ... > use Minix as the operating system that students modify and write programs to ... Now Ada compilers are availble for download, ... > PL/I available in a free download form? ...
    (comp.lang.pl1)
  • Re: Evolution of PL/I
    ... that PL/I contains, including a high speed XML parser. ... to be supplied with some compilers. ... popular languages support them. ... FAQ to make it clear that contemporary PL/I does support OOP, ...
    (comp.lang.pl1)
  • Re: Java 7 features
    ... application domain that need a log function or floating-point ... (As best I can tell from a quick Web search, ... I spent almost my entire career as a system programmer, and I remember that one of my first assignments was getting the COBOL D and FORTRAN D compilers of DOS/360 to play nicely together, for just such cases. ...
    (comp.lang.java.programmer)