[Bug testsuite/126866] rs6000: gcc.target/powerpc/mma-double-test.c fails on Power10 big-endian.

[jskumari at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126866

--- Comment #2 from Surya Kumari Jangala <jskumari at gcc dot gnu.org> ---
The behavior of __builtin_mma_assemble_pair() is correct on both LE and BE. The
testcase needs to be changed.

__builtin_mma_assemble_pair(dst, src1, src2) places src1 into the even-numbered
register of the pair and src2 into the odd-numbered register.

For the following testcase:

typedef unsigned char vec_t __attribute__((vector_size(16)));
foo (__vector_pair *dst, vec_t *src)
{
  __vector_pair pair;
  __builtin_mma_assemble_pair (&pair, src[0], src[4]);
  *dst = pair;
}

Code generated on LE:

        lxv 0,0(4)
        lxv 1,64(4)
        stxvp 0,0(3)
        blr


Code generated on BE:

        lxv 0,0(4)
        lxv 1,64(4)
        stxvp 0,0(3)
        blr

And this is as expected. The endianness is being handled correctly by the
backend.