FWIW, I use R0-R15 for32-bit  registers and RW0-RW15 for  64-bit.
On Jul 31, 2013 5:37 PM, "Paul Gilmartin" <[email protected]> wrote:

> On 2013-07-31 15:01, Gerhard Postpischil wrote:
> > On 7/31/2013 1:44 PM, Bernd Oppolzer wrote:
> >> When I came to the site where I am actually working 25 years ago,
> >> I was very surprised that they use RA to RF for 10 to 15 - I never met
> >> that before, but you get accustomed to that very quickly. It's always
> >> two characters, after all ...
> >
> > I first came across this form in the OS/360 reader/interpreter.
> > Personally I don't like it because searches on RB and RE get many more
> > false hits than R11 and R14.
> >
> And, of course, since there's no standard, the system macros
> must use pure numeric designators.  And with a mixture of
> conventions, XREF doesn't reliably list register references.
> IATYREGS is somewhat unusual among IBM macros in that it sets
> a flag so it can exit if it's used twice, and some other JES3
> macros invoke IATYREGS so they _can_ use register mnemonics.
>
> I think I've also seen GR0-GR15, which likewise reduces the
> likelihood of false hits in searches.
>
> CDC 6600 et al. made the register names predefined, reserved,
> and mandatory.  It was impossible to confuse L and LR; the
> operands distinguish them.  And no need for unintuitive
> conventions such as "7" means constant 7 and "(7)" means
> register 7 in macro arguments.  It was "7" or "X7".  So, no
> need for such as IHBINRRA.  And less incentive to print macro
> expansions; they did what they appeared to from the call.
>
> -- gil
>

Reply via email to