Re: Declaration to get 8-bit (or 16-bit) integer?

On 2007-01-30 01:03:40 -0400, nospam@xxxxxxxxxxxxx (Richard Maine) said:

glen herrmannsfeldt <gah@xxxxxxxxxxxxxxxxxxxxxxx> wrote:

Richard E Maine <nospam@xxxxxxxxxxxxx> wrote:
"Original intent"? You think that there has ever been any intent,
original or otherwise, to specify specific word sizes in any version of
the Fortran language? Apparently all the convoluted wording that goes
out of its way to express things without reference to specific sizes
must have just been an accident? The intent was to write things more
simply, but people just forgot to?

As original as it gets, for the 704 they did sign magnitude
16 bit integer arithmetic in 36 bit words. Characters were
six bits each. Bytes and 8 bit character sets hadn't been
invented yet. The card reader read cards a row at a time
into two 36 bit words, such that only 72 columns were read.

That is as close to original intent as I could come up with.

That, however, sounds like a particular implementation - not "the
Fortran language". Those are two completely different subjects. You see
any hint of evidence that the intent of "the Fortran language" was to be
designed around that machine (or any other particular machine)? I don't.
I see rather strong evidence to the contrary.

How do you tell the vendor's manual for the language from "The Language"
when there is exactly and only one vendor? And there was no language
or implementation before that.

The foible called second level definition that was in the FORTRAN 66
standard sounds like documenting the careless features of an early
optimizing compiler from the first and very dominant implementation.
And extended range DO loops. The good old days were not so good!

The Fortrans for IBM 704 (and follow-ons like 709, 7040, 7090) were very
much based on a call-by-reference semantics and IBM chose to break that
by use of copy-in/copy-out for scalars when they did the IBM/360 Fortran
compilers. My impression is that a few 360s were being delivered by 1964
but I have idea of when the first Fortran was available for them. I am
pretty sure it was usable by about 1966 but rather shakey before that.

There were lots of folks who learned assembler first and were very
unhappy campers due to copy-in/copy-out breaking the aliasing that
they had exploited in the pre-360 Fortrans. Some of it was just bad
programming but some was clever (usualy a bad word when describing
programming) use of the call-by-reference semantics.

The first ASA standard would have had the benefit of the couple 36 bit
implementaions and a 32 bit implementation so would have avoided embedding
such things. The 48 and 64 bit implementaions were to arrive soon after.
Some of the minis with 18 or 24 bits were only Fortran in name only as
they imposed no restrictions (floating point function call as subscripts!)
and had all sorts of other "features". Fortran was the high level
language around for them to imitate as assembler only was no longer
aceptable.

Was the original intent expressed by the 704 manual or by the 1966 standard?
72 columns sounds like 704 hardware but the diversity of character sizes
and implementation styles by 1966 made its effect felt in the standard.
Word and character sizes were given in the machine manual for the 704 so
were never mentioned but by the time of the 1966 standard there was a
diversity and specifics were avoided. Either way there were no 8 bit
characters.


.

Relevant Pages

  • Re: Why So Great for Numbers? / Best Compiler
    ... when dealing with code written in the Fortran language. ... They assisted development of a replacement product by DEC, purchased by Compaq>> HP>> Intel which is the current product lineage. ... Support the Original G95 Project: http://www.g95.org ...
    (comp.lang.fortran)
  • Re: [RFC] Removal of Fortran from the base system
    ... out of ignorance of the Fortran language. ... Check the ChangeLog for gfortran. ... Yes, i have no experience with Fortran, but the words high performance ... Dozens of my users use FORTRAN on our FreeBSD cluster. ...
    (freebsd-current)
  • Re: Why So Great for Numbers? / Best Compiler
    ... Kargl) wrote: ... compared to C or Lisp? ... when dealing with code written in the Fortran language. ... clear that OP lacks any interest in Fortran and is trolling. ...
    (comp.lang.fortran)
  • Re: Why So Great for Numbers? / Best Compiler
    ... Kargl) wrote: ... compared to C or Lisp? ... when dealing with code written in the Fortran language. ... clear that OP lacks any interest in Fortran and is trolling. ...
    (comp.lang.fortran)
  • Re: How to save, change the (whole) path in fortran (g95) windows
    ... outside scope of Fortran language itself... ... The original response said "not within standard Fortran". ...
    (comp.lang.fortran)