This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: powerpc-linux altivec vector function arg passing


On Wed, Apr 28, 2004 at 01:38:15PM -0400, David Edelsohn wrote:
> 	The changes to test both TARGET_ALTIVEC and TARGET_ALTIVEC_ABI
> look appropriate.

I'll pull out these changes and submit separately with a changelog entry
then.

> 	Should linux64.h always use stricter alignment for
> STARTING_FRAME_OFFSET and STACK_DYNAMIC_OFFSET, matching the darwin.h
> definitions?

Possibly.  From test cases I've looked at though, it seems that gcc gets
the alignment of vector vars correct even when STARTING_FRAME_OFFSET is
not a multiple of 16.  Or am I missing some other reason that
current_function_outgoing_args_size needs to be rounded?

Do you have any comment on the wider issue of ABI changes?

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]