Hi Carl,
On Wed, Oct 24, 2018 at 12:44:23PM -0700, Carl Love wrote:
> The scalar_cmp_exp_eq, scalar_cmp_lt, scalar_cmp_gt,
> scalar_cmp_unordered are missing support for the _ieee128 arguments.
> This patch adds the missing support and a test file for the builtins
> for both _ieee128 and double arguments.
> * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
> P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
> P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name. Add quad
> precicion entry for each overloaded builtin.
> * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
There's a stray tab in this last line.
> VSCEDPUO): Rename overloaded name.
> (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
> VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
> * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>, *xscmpexpqp): Add
> define_expand and define_insn for the instructions for the new builtins.
Make separate entries in the changelog for the expand and the insn please.
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/powerpc/float128-cmp2-runnable.c
> @@ -0,0 +1,278 @@
> +/* { dg-do run { target { powerpc*-*-* && { lp64 && p9vector_hw } } } } */
Does this need lp64? p9vector_hw implies lp64 of course, so it won't
make any difference.
> +/* { dg-require-effective-target powerpc_p9vector_ok } */
Is this not implied by p9vector_hw?
Okay for trunk with those things tuned a bit. Thanks!
Segher
