This is the mail archive of the gcc-patches@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: PING^3 for PR target/17836, PR c/10735, PR c++/16882, PR rtl-optimization/17860[3.4]


On Wed, Nov 24, 2004 at 01:41:49PM -0500, David Edelsohn wrote:
> >>>>> Janis Johnson writes:
> 
> >> The Altivec intrinsics can confuse GCC and
> >> generate incorrect warnings.
> 
> Janis> How is the ABI relevant for AltiVec builtins?  The warnings for
> Janis> passing and returning synthetic vectors from user-defined funtions
> Janis> should not apply to calls to AltiVec builtins.
> 
> 	I do not understand how your question follows from my comment.
> 
> 	The Altivec intrinsics are not actually function calls.  No
> arguments are passed and no results returned through any parameter passing
> mechanism.  Everything is within a single Altivec instruction.  If GCC is
> complaining about the intrinsic because it looks like a function call, GCC
> is confused.

That was exactly my point here and in mail I sent yesterday.  GCC is
issuing warnings about returning synthetic vectors by reference for
use of an AltiVec intrinsic when compiling with -maltivec and without
-mabi=altivec, causing lots and lots of test failures.

Janis


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