Hi Kelvin, On Fri, Jun 01, 2018 at 01:22:09PM -0500, Kelvin Nilsen wrote: > 2018-06-01 Kelvin Nilsen <kel...@gcc.gnu.org> > > * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD, > VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses. > * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove > several redundant entries.
> --- gcc/config/rs6000/rs6000-builtin.def (revision 261067) > +++ gcc/config/rs6000/rs6000-builtin.def (working copy) > @@ -1811,6 +1811,15 @@ BU_VSX_OVERLOAD_1 (VUNSIGNEDE, "vunsignede") > BU_VSX_OVERLOAD_1 (VUNSIGNEDO, "vunsignedo") > > /* VSX builtins that are handled as special cases. */ > + > + > +/* NON-TRADITIONAL BEHAVIOR HERE: Besides introducing the > + __builtin_vec_ld and __builtin_vec_st built-in functions, > + the VSX_BUILTIN_VEC_LD and VSX_BUILTIN_VEC_ST symbolic constants > + introduced below are also affiliated with the __builtin_vec_vsx_ld > + and __builtin_vec_vsx_st functions respectively. This unnatural > + binding is formed with explicit calls to the def_builtin function > + found in rs6000.c. */ > BU_VSX_OVERLOAD_X (LD, "ld") > BU_VSX_OVERLOAD_X (ST, "st") > BU_VSX_OVERLOAD_X (XL, "xl") It's not clear to me how this is different from all the other def_builtin calls, say, __builtin_vec_adde or __builtin_vec_splats. Either way, the patch is fine, okay for trunk. Thanks. Segher
