Re: The joy of FORTH (not)

In alt.folklore.computers Rich Alderson <news@alderson.users.panix.com> wrote:

Lawrence D'Oliveiro <ldo@nz.invalid> writes:
 

On Thu, 24 Oct 2024 17:13:05 -0700, Peter Flass wrote:

 

Ridiculous! 32K should be more than enough to run a decent compiler
natively.

 

It never was. Even back in Apple II days, Apple didn't recommend using
something like a Pascal compiler on their 6502-based platform.

 
My first computer was a 12K (BCD charactersw) IBM 1401 running FORTRAN IV.
Better compiler writers in those days...

Those things are not comparable.  IBM 1401 has resonably fast card
reader, many installations had tapes and/or a disc.  On home computers
frequently the only mass storage was cassete tape.  On ZX Spectrum
cassete was somewhat faster than 100 bytes per second, this is
about 1/10 of speed of 1401 card reader.  And several home computers
had slower cassete interface than Spectrum.  So normal tactic
on home computers was to keep everthing in RAM and use one-pass
compiler.  And Pascal on ZX Spectrum needed 25 kB RAM.  Compilers
on 1401 were multi-pass and used external storage.

Also, we expect progressively more from compilers.  As a student
I started with ICL Fortran.  But in the next semester the course
moved to IBM 360 compatibles and IBM Fortran.  Suddenly things
that did work in ICL Fortran no longer worked, in particular IBM
allowed only very simple expressions as array indices, while ICL
allowed arbitrary expressins.  Early compiler allowed only very
short varisble/subroutine names, forcing use of obscure names.

Instruction set also have influence: 1401 looks easy to program,
but instructions tend to be somewhat long.  Z80 instructions are
shorter, but to do anything one needs many istructions, so program
actually tend to be longer.  On Z80 something like p-code may
be attractive to make programs smaller.

--
                              Waldek Hebisch