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: [patch] fix red zone altivec saving



On Mar 25, 2007, at 10:29 PM, Andrew Pinski wrote:


On 3/25/07, Eric Christopher <echristo@apple.com> wrote:
So, it turns out that the red zone on ppc doesn't include the altivec
registers. OK, no problem on those targets (ABI_V4) that happen to
push the stack early enough. Unfortunately darwin isn't one of them
so a little rearrangement is necessary. However, this doesn't really
affect V4 much in that we just save the registers a little later in
the sequence after we've pushed.

Tested by bootstrap and regression test on ppc-darwin. OK?

Is this true for say PPC64-linux also which does have the red zone. Also ppc-aix which does support VMX and has the red zone (though GCC does not support VMX there yet).

I really think this should be a darwin specific patch if both the
PPC64-linux ABI (which you can find easily) and ppc-aix's red zone
supports VMX.

None of the zones (looking at the definition of push_p) are large enough to support the altivec registers.

-eric


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