Konrad Schwarz (no.spam.delete@this.no.spam) on January 13, 2020 3:49 am wrote:
> Adrian (a.delete@this.acm.org) on January 11, 2020 6:29 am wrote:
> > The method used by DEC PDP-11 to extend the address space from 16 bit to 22 bit was
> > much less convenient than the segmentation of 8086 or of 80286, so at introduction
> > nobody criticized 8086 for its segmentation, because it was progress anyway.
>
> Of course this gave rise to the pipe, Unix, and software tools.
>


I am not sure what you meant by that.

Because swapping memory pages on PDP-11 was much more difficult than using far pointers on 8086, most PDP-11 programs were limited in size to 64 kB of code + 64 kB of data, like in the small memory model of IBM PC.

Even if you had 4 MB of DRAM in your PDP-11, the individual applications had to be much smaller and the large memory only allowed you to execute many processes in parallel.

Programs that could not fit in 64 + 64 had to use "memory overlays", i.e. the program had to be divided in sections loaded on demand and then deleted, e.g. for a multi-pass compiler.


So you might have a point after all, because such an environment would encourage the implementation of large data processing tasks by dividing them into many small processes communicating between themselves, e.g. by pipes, as this could use efficiently the entire available memory, unlike a monolithic program.