[Bug c/59854] New: Types with specific endianness

[Bug c/59854] New: Types with specific endianness

[hpa at zytor dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59854

            Bug ID: 59854
           Summary: Types with specific endianness
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: enhancement
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: hpa at zytor dot com

For integer (and ideally floating-point) types, it would be very nice to be
able to explicitly specify bigendian or littleendian, with the definition that
when stored in memory, these types are byteswapped if necessary.  This would
allow a much more natural and less error-prone way to build data structures for
interchange than what is currently possible.

Obviously, the "as if" rule applies: if it is not observable that something is
stored in memory (e.g. a stack variable for which an address is never taken)
the compiler is allowed to elide the byte swapping as an optimization.

Ideally, this should be added to the C and C++ standards, but perhaps an
implementation in gcc would help make that actually happen.

[Bug c/59854] Types with specific endianness

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59854

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2014-01-17
                 CC|                            |ebotcazou at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #1 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
There is an implementation available on the scalar-storage-order branch.