[email protected] (Gerhard Postpischil) writes:
> I think you misread his message, which started with command
> chaining. But I would be interested in which control units and
> controllers allow data chaining beyond 65KiB.

re:
http://www.garlic.com/~lynn/2013o.html#6 hexadecimal

note a lot of original data chaining was for non-contiguous storage
transfers (scatter/gather). cp67 doing emulated i/o had to use it for
shadow ccws involving (non-contiguous) page crossing. channel
architecture proscribes that ccws aren't prefetched (executed serially
one at a time, even datachaining) ... and their could be timing
difficulties with cp67 emulated i/o ... when the native
(non-datachained) i/o ran find.

the issue of page non-contigious i/o was addressed in 370 with IDALs ...
where IDAL channel architecture allowed the the addresses to be
pre-fetched. IDALs was also used later as part of hack adding more than
16mbytes to 3033 ... even tho instructions and CCWs couldn't address
more than 16mbytes.

note that many of the i/o devices are now industry standard with IBM
legacy controllers and devices (like CKD) purely a simulated
artifact. They aren't restricted to record lengths imposed by IBM's
legacy CCW architecture.

-- 
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