[email protected] (Paul Gilmartin) writes:
> Decades ago, I had experimented with similar techniques, even so far as
> paired +- RLDs.  In the course of that, I discovered that CMS obeys its
> own rules.  If I had unresolved WXTRN + absolute offset, it got relocated
> anyway because CMS assumed the nonzero result meant the WXTRN was
> resolved.

re:
http://www.garlic.com/~lynn/2016h.html#98 A Christmassy PL/I tale
http://www.garlic.com/~lynn/2016h.html#101 Multitasking question

part of the issue was that CMS had a 64kbyte os/360 system services
simulation implementation. With the introduction of MVS ... and its
8mbyte kernel image ... there was a joke that CMS 64kbyte os/360 system
services simulation was significantly more effective than the MVS 8mbyte
os/360 system services simulation.

In the mid-70s, somebody in the burlington vm370/cms development group
did a rewrite increasing the size to approx 128kbyte os/360 system
services simulation ... with lots more feature/function compatibility
(including read/write os/360 disks supporting file access methods).

However this was about the time that Future System was failing and the
mad rush to get stuff back into the 370 product pipelines ...  including
kicking off 3033, 3081/xa and mvs/xa. the head of POK tells corporate
that he needs vm370 product killed and all the people moved to POK or
otherwise he won't be able to ship mvs/xa on schedule (this is still
mid-70s). All the enhancements not already shipped ... disappear in the
shutdown of the burlington site.
http://www.garlic.com/~lynn/submain.html#futuresys

They were not planning on telling the burlington group until shortly
before the shutdown goes into effect, to minimize the number of people
that could escape the move. Unfortunately(?) the shutdown leaks several
months early and lots manage to escape (there was joke that the head of
POK is one of the major contributors to DEC vax/vms) ... and then there
was witchhunt to find who leaked the information ... fortunately, nobody
gave me up.

Endicott finally managed to save the vm370 product mission ... but they
had to reconstitute a development group from scratch ... and Endicott
was much more interested in providing dos/vs system service simulation
(than os/360 system services simulation)

lots of past post about location independent code and fighting RLDs
http://www.garlic.com/~lynn/submain.html#adcon

other multitasking tivia: charlie had invented compare&swap (chosen
because CAS are his initials) ... while doing fine-grain cp67
multiprocessor locking at science center ... some posts
http://www.garlic.com/~lynn/subtopic.html#545tech
and
http://www.garlic.com/~lynn/subtopic.html#smp

attempting to get compre&swap added to 370 was repulsed because the POK
favorite son operating system people said that test&set (from 360/65MP)
was more than sufficient. 370 architecture owners said that in order to
add compare&swap to 370, uses other than SMP kernel locking was
needed. Thus was born the application multitasking/multithreaded
examples that still are included in appendix of POO. This was picked by
large mulithreaded/multitasking subsystems like DBMSes (IMS, DB2, etc).

In the 80s, other platforms started adding compare&swap instructions (or
instructions with semantics similar to compare&swap) ... as part of
supporting large commercial DBMS. Platforms w/o compare&swap semantics
had to use kernel calls with significantly higher overhead.

-- 
virtualization experience starting Jan1968, online at home since Mar1970

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

Reply via email to