Re: [TESTCASE] gcc miscompiles Altivec code

[Aldy Hernandez <aldyh at redhat dot com>]

On Tuesday, February 26, 2002, at 09:15  AM, Daniel Egger wrote:

> Hija,
>
> I distilled a (not so) small example for code being miscompiled
> by gcc from yesterday on linux-ppc.
> This one will segfault on my maschine when compiled with -O0.

this is really helpful.  thanks.

however...

tried it on my darwin box... compiles fine with -O[012].  i tried
it on apple's altivec compiler and the resulting vectors are all zero.

i added a loop like this to print them:

   for (i=0; i < 7; ++i) {
         printv((vector signed int)vec[i]);
         printv((vector signed int)v1[i]);
         printv((vector signed int)v2[i]);
   }

typedef vector signed int vecint;

void printv(vecint v)
{
         int foo[4] __attribute__((aligned(16))) ;

         vec_st (v, 0, foo);
         printf("%x %x %x %x a3\n", foo[0], foo[1], foo[2], foo[3]);
}

i ran with cvs gcc, and it not only compiles but yields the same 0's
as the apple compiler.

are you passing any special flags to cc1?