Re: How does Alsys index bit-arrays on Intel ?

[bashford@srs.loral.com (Dave Bashford)]

In article <1993Mar23.202742.26997@mksol.dseg.ti.com> strohm@mksol.dseg.ti.com (john r strohm) writes:
>In article <C4CLou.I5z@watson.ibm.com> ncohen@watson.ibm.com describes an
>Ada algorithm to determine which bit (most or least significant) is numbered
>how.
>
>The problem is that this tells only how that particular version of that
>particular compiler for that particular host and target numbers them.  This
>does NOT give any assurance that a compiler from a competing vendor will
>do the same thing, nor does it even cover the case of another compiler from
>the same vendor, for the same host but a different target, or even (in a
>REALLY baroque scenario) for a subsequent release of the SAME compiler.
>The language of the standard allows the vendor to do what he/she wishes.
>
>This, in my personal opinion, is one of the few places where Ada goofs.
>
>Now, the flip side is that there is a good argument for making it
>implementation-dependent.  Some machines still number bits left-to-right,
>while others number right-to-left, and we still have Endian problems
>every time we change processors (almost).  Choosing one scheme and
>enforcing it is going to cause major heartburn for SOMEBODY, as they
>curse the standard because none of their databooks are useable: they
>all have the bits numbered backwards.
>

Would it be a "major heartburn for SOMEBODY" ? Both Intel and Motorola
number the bits within a byte right-to-left. The 68K intruction set has
a bit-set instruction with right-to-left ordering; to do the same thing
on the Intel requires masking and or'ing. Yet the Ada compiler for the
68K seems to do just fine translating bit numbers. I don't see where it
would be difficult to write a compiler either way - what is a _pain_ is
trying to write portable code in such an environment ! I was under the
apparently mistaken assumtion that portability (platform independence)
was one of the main intentions of the DoD when they decided to create/use
a common language.

This, in my personal opinion, is one of several places where Ada
goofs.  As someone who writes a lot of reusable subsystems, I find all
of the little implementation-dependent features a real pain. A few
things that would help enormously are a standard pre-precessor, a
well-defined alternate-body mechanism, and/or a standard Make tool. I
look with envy at all of the PD stuff on the network that will compile
and run on everything from a DOS to Unix and even VMS, yet I can't
write Ada source that will compile the same on Sun and HP.

P.S.	I hope this doesn't sound like the start of a religious
	argument - I did delete some of the stronger stuff :-) I'm not
	religious just frustrated.  I would like to see a useful thread
	dealing with these issues.
-- 

db
bashford@srs.loral.com (Dave Bashford, Sunnyvale, CA)