On Tue, Oct 26, 2021 at 7:16 AM Peter Relson <[email protected]> wrote:

> John M wrote:
> <snip>
> I really would like the Linkage Stack to become generally
> useful. So I could just use a BAKR to save state on entry and PC to
> return.
> </snip>
>
> In what way does BAKR...PR not let you accomplish that (if by "state" you
> are referring to thing such as registers and PSW state)?
> BAKR 14,0
> ...
> PR
>

The problem that I am thinking of is that an interrupt occurs if the stack
is exhausted before everything is saved. The last time I looked,
this causes an ABEND. IMO to be really useful z/OS would need to
incorporate some way to "automatically" expand the size of the Linkage
Stack somehow, rather than forcing each program to implement this function.
Perhaps a way to "chain" individual Linkage Stack areas together. A "stack
overflow" would cause z/OS to create a new Linkage Stack and chain it to
the previous and automatically retry, or ABEND if this operation failed
somehow. On an "underflow", z/OS wold see if there was a previous Linkage
Stack, free the current one, make the previous one the active one and retry.



>
> with this sometimes being preceded by BSM 14,0 in order to make sure that
> the reg 14 is properly defined for the AMODE needed (unnecessary if you
> got here by something like BASR 14,15 or BRASL 14,xxx).
>
> As pointed out, this would still be considerably slower than the analogous
> register saving/restoring.
>

and that's the other part. I don't know why this is "slow". But perhaps the
hardware people could optimize it more, or add new "assist" circuitry. I
realize this is costly & is likely why it won't be done. The save areas
work, so why invest money in this rather than elsewhere that might be more
helpful.

The world is not perfect. FSVO "perfect".



>
> Peter Relson
> z/OS Core Technology Design
>
>

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

Reply via email to