From: Robert A Duff <bobduff@shell01.TheWorld.com>
Subject: Re: Efficient Matrix?
Date: Sat, 28 Dec 2002 23:42:33 GMT
Date: 2002-12-28T23:42:33+00:00 [thread overview]
Message-ID: <wccd6nld82e.fsf@shell01.TheWorld.com> (raw)
In-Reply-To: slrnb0s84c.9ge.adi@drcomp.erfurt.thur.de
Adrian Knoth <adi@drcomp.erfurt.thur.de> writes:
> Robert A Duff <bobduff@shell01.TheWorld.com> wrote:
>
> > turns it into a binding requirement. 13.1(20) also mentions this.
>
> ACK.
>
> > So a standards conforming compiler can get away without supporting any
> > representation items at all. But a compiler that claims to support the
> > SP annex must obey 13.2(9).
>
> 9 For a packed array type, if the component subtype's Size is less
> than or equal to the word size, and Component_Size is not
> specified for the type, Component_Size should be less than or
> equal to the Size of the component subtype, rounded up to the
> nearest factor of the word size.
>
> How does this _forces_ a compiler to set a packed boolean array's
> Component_Size to 1? As I understand the paragraph it determines
> the compiler's behaviour if the Component_Size of a packed array is
> _not_ defined by the user.
I guess I was talking about that case. E.g:
type T is array (Positive range <>) of Boolean;
pragma Pack(T);
The component subtype is Boolean.
Boolean'Size = 1, by 13.3(49). Also 13.3(55), combined with the SP
annex thing.
This size (1 bit) is <= word size on every machine. And it's a factor
of the word size on every machine. So the compiler must choose
T'Component_Size <= 1. And it can't very well choose < 1, so it must
choose = 1.
If you specify T'Component_Size, then the above paragraph is saying that
overrides the pragma. So if you say:
type T is array (Positive range <>) of Boolean;
pragma Pack(T);
for T'Component_Size use 2;
then T'Component_Size = 2. This is required to be supported by
13.3(73), combined with the SP annex thing, assuming the word size
is an even number of bits (which is quite common ;-)).
> I guess I'll send a bug-report. Unfortunately the gcc-bug-database is
> write-only at the moment, so to say, there seems to be nobody carrying
> much about ProblemReports.
Unless you're a paying customer, they have no obligation to fix things
soon. But they do seem to fix things "eventually".
- Bob
next prev parent reply other threads:[~2002-12-28 23:42 UTC|newest]
Thread overview: 20+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <3e0b2a66_4@news.bluewin.ch>
2002-12-26 22:09 ` Efficient Matrix? Adrian Knoth
2002-12-27 0:23 ` Alvery Grazebrook
2002-12-27 9:53 ` Adrian Knoth
2002-12-27 16:58 ` Robert A Duff
2002-12-28 1:44 ` Nick Roberts
2002-12-28 13:00 ` Adrian Knoth
2002-12-28 15:21 ` Bill Findlay
2002-12-28 15:48 ` Adrian Knoth
2002-12-28 23:26 ` Adrian Knoth
2002-12-28 16:07 ` Robert A Duff
2002-12-28 17:25 ` Bill Findlay
2002-12-28 17:35 ` Bill Findlay
2002-12-28 20:51 ` Robert A Duff
2002-12-28 15:58 ` Robert A Duff
2002-12-28 17:19 ` Adrian Knoth
2002-12-28 19:16 ` James S. Rogers
2002-12-28 20:45 ` Robert A Duff
2002-12-28 22:07 ` Adrian Knoth
2002-12-28 23:42 ` Robert A Duff [this message]
2002-12-27 12:25 ` Gautier
replies disabled
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox