On 5/09/2026 5:42 am, Phil Smith III wrote:
If the idea for ARM on Z is to allow ARM binaries to run alongside Z binaries,
how does that work with Big and Little Endian? Or will there be Middle Endian?*
I'm not sure that big/little endian is a big deal. It affects the
behavior of a few instructions, but you're running a different
instruction set anyway. More interesting is the question of how the
system knows whether this sequence of 1s and 0s is Z instructions or
ARM. I guess it could work similarly to AMODE with an explicit switch.
My question is why? Are there ARM programs out there that people want to
run on Z but can't recompile? I doubt it. OS functions are usually the
problem rather than the machine instructions.
<speculation>
IBM want to stop making specialized CPUs and switch to ARM processors.
However there *is* a lot of z/OS software that can't be recompiled for
another platform. Particularly assembler but other stuff as well. How
might it work? Switch to ARM processors. Implement Z instructions in
microcode. For maximum performance, recompile programs targeting ARM.
Otherwise, z/OS instructions are handled by microcode (many probably map
1:1 to an ARM instruction). For systems running processors at less than
full speed, it's probably not hard to tweak the dial to give the same
performance as current systems...
</speculation>
--
Andrew Rowley
Black Hill Software
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN