-m(no-)align-double compatible libgfortran
Erik Schnetter
schnetter@cct.lsu.edu
Wed Feb 21 13:39:00 GMT 2007
On Feb 21, 2007, at 00:28:35, Jerry DeLisle wrote:
> Erik Schnetter wrote:
>> On Feb 20, 2007, at 13:15:39, Thomas Koenig wrote:
>>> On Tue, Feb 20, 2007 at 11:47:59AM -0600, Erik Schnetter wrote:
>>>
>>>> I have a patch that works, but Richard Günther didn't like it
>>>> because
>>>> it is not "robust".
>>>
>>> Could we use a pointer to a double, instead?
>> I guess that would be possible.
>>> And where is this double lurking, anyway? I missed that :-)
>> It is a 64 bit integer, not a double. -malign-double also affects
>> these. It is used to mark a position in a file.
>> -erik
>> --Erik Schnetter <schnetter@cct.lsu.edu>
>> My email is as private as my paper mail. I therefore support
>> encrypting
>> and signing email messages. Get my PGP key from www.keyserver.net.
> Erik,
>
> Can you send the patch to me to study. I would also like to take a
> look at feasibility of changing to a pointer.
Jerry,
I sent the patch to <http://gcc.gnu.org/ml/gcc-patches/2007-01/
msg02223.html>. This patch exchanges some structure members; it does
not add padding. The problem is with the entry "GFC_IO_INT rec",
since intio is a 64 bit integer.
I don't know where the entry rec is used, but it seems to me that
changing it to a pointer should be possible. I assume that the
compiler allocates such a structure on the stack when Fortran calls I/
O routines; in this case, you would need to allocate the structure
and the GFC_IO_INT and link them appropriately, and adapt the code
that accesses this field.
-erik
--
Erik Schnetter <schnetter@cct.lsu.edu>
My email is as private as my paper mail. I therefore support encrypting
and signing email messages. Get my PGP key from www.keyserver.net.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: PGP.sig
Type: application/pgp-signature
Size: 186 bytes
Desc: This is a digitally signed message part
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20070221/6d5aa32e/attachment.sig>
More information about the Fortran
mailing list