From: muta...@gmail.com
On Tuesday, September 13, 2022 at 10:16:55 AM UTC+8, muta...@gmail.com wrote:
> On Tuesday, September 13, 2022 at 12:24:43 AM UTC+8, anti...@m
th.uni.wroc.pl wrote:
> > muta...@gmail.com wrote:
> > > On Monday, September 12, 2022 at 4:23:43 AM UTC+8, s_dub...@yahoo.com
wrote:
> > > > On Sunday, September 4, 2022 at 1:55:32 PM UTC-5, muta...@gmail.com
wrote:
> > > > > On Monday, September 5, 2022 at 1:03:24 AM UTC+8, Scott Lurndal
wrote:
> > > > > > "muta...@gmail.com" writes:
> > > > > > >On Sunday, September 4, 2022 at 9:25:34 PM UTC+8, wolfgang kern
wrote:
> > > > > > >> On 04/09/2022 06:32, muta...@gmail.com wrote:
> > > > > >
> > > > > > >Noone had done segmentation before and it
> > > > > > >would take decades of science still to
> > > > > > >work through the repercussions.
> > > > > > Several very successful systems that predate the
> > > > > > 8086 by a full decade used segmentation; examples include the
PDP-11,
> > > > > > the B6500 et alia.
> > > > > They presumably didn't do 4 bit shifts or similar
> > > > > so didn't hit the a20 issue nor the flexible shift issue.
> > > > >
> > > > > The a20 is probably because of cp/m PSP though,
> > > > > not segment shift.
> > > > They presumably didn't do 4 bit shifts or similar
> > > > so didn't hit the a20 issue nor the flexible shift issue.
> > > >
> > > > The a20 is probably because of cp/m PSP though,
> > > > not segment shift.
> > > > ~~~
> > > >
> > > > Gosh, istm you have these concepts jumbled up.
> > > >
> > > > Sorry for the review but..
> >
> > >
> > > So first we have the justification for address wrap.
> > > So that call 5 could be implemented.
> > >
> > > Perhaps we need to go back to the 8080
> > > and/or cp/m to see whether the call 5
> > > could have been replaced with something
> > > that could have future-proofed the
> > > situation.
> > No "perhaps". Call 5 was for CPM compatibility. Original
> > CPM was for 64kB address space, so no problem _preserving_
> > compatiblity with respect to segment manipulation/wraparound:
> > this was new stuff. 8086 was different instruction set
> > so if was possible to implement equivalent functionality
> > in different way.
> >
> > Even if you insist on having exactly the same functionality,
> > in PSP it is easy to do without wraparound: just put extra far
> > jump at end of BIOS ROM and in PSP have jump to this extra jump.
> > Cost is extra 4 bytes of ROM used by the jump and extra
> > machine cycles to execute extra far jump per each call 5.
> >
> > Of course, once BIOS without this extra call was in the wild
> > DOS had to support such BIOS-es and gate A20 was logical
> > solution...
> >
> > --
> > Waldek Hebisch
> Even if gate a20 was a logical solution,
> would it have been reasonable for msdos to
> have an explicit bios call to enable wraparound?
>
> Even if existing bioses don't implement that
> call, it doesn't matter, because the 8086 is
> going to wrap around anyway.
>
> It's only when the 80286 arrives that the
> bios call needs to actually work
Actually, there could be one bios call to query
whether the bios has those dedicated 4 bytes.
If that returns success, then that is what
is used for call 5.
Otherwise, a bios call is done to disable a20.
Unless you are on a machine with more than 1 MB
of memory, in which case the high memory
area is populated with those 4 bytes.
Actually, why is disabling a20 ever
needed? Won't you always be in a position
to populate high memory if it exists,
and if it doesn't, there's no issue anyway.
This is why I like to discuss things.
I can't derive anything in a vacuum.
--- SoupGate-Win32 v1.05
* Origin: you cannot sedate... all the things you hate (1:229/2)
|
