Glenn asks:
> Stan writes:
> > I forgot to add...the other major difference is that NM code can address
> > a heck of a lot of code/data: many, many gigabytes (1 to 4 gigabytes easily,
> > and thousands of 4 gigabyte chunks with slightly more work ... far far far
> > more than HP-UX (or most any other machine)).
>
> But...but...but...How can machine instructions -- the same instructions
> used both by MPE and by HP-UX -- be compared to an OS? Is this really
> just saying that HP-UX does not take full advantage of the underlying
> hardware??
HP-UX *barely* uses any of the addressing capability of PA-RISC.
MPE/iX uses all of it!
With MPE/iX, you can HPFOPEN/longmapped a thousand files, each of
4 GB. Since you get back a virtual address for each file, that gives
you 4 TB of data addressability! (Admittedly, segmented into 4 GB
chunks, and admittedly somewhat difficult to use from languages that
don't support 64-bit pointers (COBOL/iX, FORTRAN/iX, gcc, RPG/iX,
BusinessBASIC/iX) (Said differently, only HP C, Pascal/iX, and SPLash!
can handle 64-bit pointers natively.)
HP-UX roughly looks like a 2-GB system space and a 2-GB process local
space ... sort of like a VAX ... and that's *all*. You can't get
access to other spaces.
HP, of course, has never tried to use this addressing capability as
a competitive advantage. After all, it might take away sales from HP-UX :)
HP has announced they're working on files > 4 GB. One would assume that
when that ships, you'll be able to have many "large files" open at a
time...giving you many chunks of linear address space much larger than
4 GB!
Now...if only I had a TB of disk storage! (and a way to back it up :)
--
Stan Sieler [log in to unmask]
http://www.allegro.com/sieler/
|