This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Selecting the correct default ABI for the iWMMXt processor
- From: Nick Clifton <nickc at redhat dot com>
- To: Ian Lance Taylor <ian at wasabisystems dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Fri, 16 Apr 2004 11:47:24 +0100
- Subject: Re: Selecting the correct default ABI for the iWMMXt processor
- References: <407D7459.9050308@redhat.com> <m34qrmjq57.fsf@gossamer.airs.com>
Hi Ian,
It seems very strange to me that a -mcpu option will change the ABI.
It won't, It will select an ABI to use if the user has not specified
one. The code in arm_override_options() already makes a decision of
which ABI to use when the user has not specified one. This patch simply
extends the logic used so that the ABI selected will be compatible with
the iWMMXt processor.
Also, it appears that your patch will force the use of ARM_ABI_AAPCS
(why not ARM_ABI_IWMMXT?)
Oops - that was a typo which I will fix once we have sorted out the rest
of this patch.
Also, I think that changing
! MULTILIB_OPTIONS = mlittle-endian/mbig-endian
to
! MULTILIB_OPTIONS = mbig-endian
is definitely wrong. You should use MULTILIB_DEFAULTS for this kind
of thing. The same for your other MULTILIB_OPTIONS changes.
Why is it wrong ? It works. It eliminates a duplicate set of multlibs
and it keeps the makefile fragment simple.
Cheers
Nick