[boost] Re: Raw Array Classes

From: Ed Brey (brey_at_[hidden])
Date: 1999-08-20 12:42:06

Email seemed somewhat boring, until Reid Sweatman wrote:
>
>> defect report requiring the vector be contiguous.  I'm not sure about
>> data alignment issues, though, but I expect that the the consideration
>> is for it to be the same as an array.
>
>What do you mean by "same as an array?"  The items in an array could have
>widely varying alignments, depending on whether you specified a packing size
>or an alignment boundary for data items.  For instance, specifying an even
>alignment is just about guaranteed to create differently-sized array
>elements than specifying a 16-byte alignment, depending, of course, on
>what's in the array.  Do you mean that whatever alignment is current when
>the array is allocated determines its start address?  What about the start
>addresses of its elements?

The defect report can give you the details of the proposed solution much
better than I can.  See:

http://anubis.dkuug.dk/jtc1/sc22/wg21/docs/lwg-issues.html#69

From the requirements in the proposed solution, you can deduce that
the elements would be stored "like an array", for a given packing.
I don't think that C++ addresses how to deal with varying packing schemes
within the same program or translation unit.  That is left to how the
implementation wants to implement its pragmas and compiler switches.
I would expect that implementations will apply the current packing that
is in place when the vector is defined.