Oppolzer - Informatik / Blog

Am 03.01.2012 00:12, schrieb P.G.:
> On Jan 2, 2012, at 03:11, Bernd Oppolzer wrote:
>> On Sun Solaris, there were other kinds of errors, which had to do with the alignment of shorts,
>> ints, and longs. While optional on other platforms, this was mandatory on Solaris.
>>
> This could also be a consequence of presuming little-endian.

we have the endian-issues with z Arch, so we were accustomed to this; no
problems so far. But what was really new when we started with Solaris, were
runtime errors, when pointers to long variables didn't contain addresses which
are multiples of four. This happened with S/360, but never since ...

>> We are also constantly fighting with rounding problems - successfully most of the time - ,
>> because we use the classical HEXADEC representation for floating point on z/OS
>> (we have to communicate with PL/1 and - classical - ASSEMBLER).
>>
> Are you suggesting that ASSEMBLER lacks support for either IEEE BFP or DFP?
> Don't know about PL/I.

no, I know that the z Arch supports all that. But we started our math package in
1993 ca., and at that time HFP was a natural choice on C/370 (don't recall if
there was an alternative at that time). Later, we didn't want to change, simply
because we coped with the rounding problems. Well, that's not entirely true: for
some customers, we deliver z Arch versions with IEEE (= BFP).

PL/1 supports today BFP etc, too.

> -- G.
>