This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: altivec patches
- To: Aldy Hernandez <aldyh at redhat dot com>
- Subject: Re: altivec patches
- From: David Edelsohn <dje at watson dot ibm dot com>
- Date: Wed, 07 Nov 2001 13:12:55 -0500
- cc: gcc patches <gcc-patches at gcc dot gnu dot org>
>>>>> Aldy Hernandez writes:
Aldy> i think rth was talking about the stack boundary when he suggested
Aldy> things needed to be abi dependent. right now i have left biggest
Aldy> alignment to be 128 regardless of ABI.
>> I don't disagree with making BIGGEST_ALIGNMENT 128, but some AIX
>> and Darwin macros use BIGGEST_ALIGNMENT and expect it to be 64. If you
>> change BIGGEST_ALIGNMENT, the other macros in rs6000 directory need to use
>> 64 or some other symbolic name for 64.
Aldy> we could change them all to 128. or if people are paranoid we could set
Aldy> it to:
Aldy> (TARGET_ALTIVEC ? 128 : 64)
Aldy> which would be dependent on -maltivec, and NOT -mabi=altivec.
Aldy> basically, altivec instructions with current abi.
aix.h, darwin.h, and linux64.h ROUND_TYPE_ALIGN macro assumes
BIGGEST_ALIGNMENT is 64. I thought that Apple already had changed this,
but they never submitted a patch. If you change BIGGEST_ALIGNMENT to 128,
you need to fix those macros.
David