On Tue, Mar 28, 2017 at 12:30 PM, Paul Gilmartin <0000000433f07816-dmarc-
[email protected]> wrote:

> On Tue, 28 Mar 2017 13:04:10 -0400, Steve Smith wrote:
> >
> >I presume IBM would like to deprecate and retire dataspaces completely,
> >although the eternal backward-compatibility may prevent that for 100 years
> >or more.
> >
> (Wow!  We're halfway there!)
>
> My impression, also.  Is there anytning possible with AR-mode that can't be
> done with AMODE 64?  (May I assume that AR mode-mode and AMODE 64
> are mutually exclusive?)
>

​Why assume that?  I see no reason for it, hardware wise.

One thing that can be done with the use of multiple data spaces is a form
of "overwrite protection". Imagine that for a set of "objects", you create
a data space to contain them. The "pointer" would consist of the
ALET+address. Now, for a different set of objects, you use a separate data
space, with (of course) a different ALET. Code which loads the ALET for
dataspace1 cannot harm any data in dataspace2. As opposed to "intermixing"
the objects' data in a single AMODE(64) area.​



>
> What's the peformance consequence of AR-mode?  I might expect it to be
> slower than AMODE 31 but faster than AMODE 64.
>
> I feel similarly about page protection keys.  Segment protection should
> make
> them obsolete.
>

​Only if you completely rearchitect how z/OS works. z/OS expects some
things to be in "shared" memory (PSA, NUCLEUS, [E]LPA, [E]CSA​). I'm fairly
sure that PSA will always need to be shared and so protected. Now, if z/OS
went to a true micro-kernel like MACH is said to have, then the use of
protection keys could be more limited. But at a very high, relative, CPU
cost. Hardware key protection is fairly cheap compared to what would likely
be a PC-SS instruction for a MACH-like kernel. And, as IBMers tend to ask
"What is the business advantage? Who will pay for it? What are you willing
to give up or delay while we do this?"  If you really want such a system,
look at the IBMi, which runs on Power8 with no hardware keys and a single
address space which encompasses both what we think of as "memory" and the
entire "permanent" file system. It's a very interesting system.



>
> -- gil
>
>
-- 
"Irrigation of the land with seawater desalinated by fusion power is
ancient. It's called 'rain'." -- Michael McClary, in alt.fusion

Maranatha! <><
John McKown

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to