On 14/03/2020 22:03, Martin Gregorie wrote:
> On Sat, 14 Mar 2020 13:59:11 +0000, mm0fmf wrote:
>
>> On 13/03/2020 23:45, Michael J. Mahon wrote:
>>> The Natural Philosopher  wrote:
>>>> On 13/03/2020 14:32, Martin Gregorie wrote:
>>>>> But I thought then, and still think now that its fundamental flaw is
>>>>> that its designers
>>>>
>>>> Its designer. Grace Hopper, a US Navy rear admiral.
>>>>
>>>>> thought that there are two types of programmer: those who understood
>>>>> Fortran/Algol/C style assignments and others who could only
>>>>> understand something that looked like English text, but who could
>>>>> nonetheless understand all the ramifications of a complex data
>>>>> declaration or some of the more arcane variations of a PERFORM or a
>>>>> SORT statement. Thats' false thinking, of course, but it did leave us
>>>>> with the most verbose computer language in the known universe.
>>>>
>>>> No,that is assembler Grace wrote Cobol for accountants who understood
>>>> their data, not computers.
>>>>
>>>> It was also written so that it could run on small RAM  computers A
>>>> considerable achievement. I regard it with teh same awe as 'C' in that
>>>> it allowed massive programs to be written by quite ordinary
>>>> programmers
>>>>
>>>>
>>> Just about any language can be compiled on a small computer as long as:
>>>
>>> 1) you have a couple of tapes,
>>> 2) you are willing to sort the source program multiple times, and 3)
>>> time is not an object.
>>>
>>> In fact, this was the common compiler approach!
>>>
>> It's a long, long, long time since I tried it (36 years ago) but ISTR
>> compiling an ALGOL program on my DG Nova 1200 was an 11 pass operation
>> using paper tape. i.e. 11 tapes in the compiler/assembler/linker/loader
>> plus the intermediate tapes punched.
>
> The old Elliott 503 I learnt Algol 60 on, in 1967, might have had paper
> tape input and paper tape plus lineprinter output and only 8 Kwords of 39
> bit memory, but it also has a speed boost secret -  32 Kwords of 50 uS
> 39bit core backing store.
>
> No linker needed because separately compiled programs could call each
> other - effectively dynamically linked libraries in the late 60s!
>
> The compiler, assembler and utilities plus any big data arrays that might
> belong to user programs and (presumably) compiler workfiles all lived on
> the core backing store, so the thing was quite quick.
>
> You could book blocks of time for serious stuff, or run compiles and
> small jobs in automatically timed 3 minute slots. That was plenty for my
> work: I was using the beast to analyse paper tape output from a Mossbauer
> Spectrometer. A compile and analytic run on my data took well under 3
> minutes (no storage space for compiled programs was available), so I
> never needed to book time on it.
>
> A lovely, but very large, box. It was entirely built using discrete
> transistors and ferrite core memory since it predated the invention of
> the first integrated circuits. Its floating point operations were
> slightly faster than its integer arithmetic, and no, I have no idea how
> its designers managed that feat.
>
>

My NOVA that lived in the garage had some DG cassette tape drives that
didn't work, an 8 port RS232 card, and 16kW of ferrite core. The CPU was
all TTL and the ALU was bit slice, 1x 74181 TTL ALU unit with  each
16bit word split into 4bit nibble operations. It took 4 clocks to do
each instruction. ISTR the NOVA 3 was faster as it had 4x 74181 in it.
And a tape reader.

Classic 'proper' computing were you toggled in the boot loader on the
switches, loaded the tape and hit go and off it went.

The instruction set was wonderfully simple to use in assembler. No
hardware stack, you saved the return address for a subroutine in the
first few words of your called function as there was no ROM to worry
about then.

It had been scrapped a few times by the time I got it. Just a toy but
one that was too big for the house.

--- SoupGate-Win32 v1.05