On Tue, Oct 18, 2011 at 22:21, Richard Henderson <r...@twiddle.net> wrote: > On 10/18/2011 01:24 PM, Blue Swirl wrote: >>> #ifdef TARGET_SPARC64 >>> -float64 helper_fabsd(CPUState *env, float64 src) >>> +float64 helper_fabsd(float64 src) >> >> This probably should go to previous patch. > > Sure. > >>> +/* Turn off the stupid always-inline hack in osdep.h. This gets in the >>> + way of the callback mechanisms we use in this file, generating warnings >>> + for always-inline functions called indirectly. */ >>> +#define always_inline inline >> >> It would be better to just delete the offending (or all) inlines. > > I certainly would like to delete the offending hack in osdep.h. > > The inline markers themselves are generated by def-helper.h, and are required > so that we don't wind up with a corresponding number of defined-but-not-used > errors from the helper.h definitions. > > I really didn't know any one way to handle this situation that would be > immediately acceptable to everyone. I assumed limiting the change to > the sparc front-end would minimize the pushback.
It should also be possible to add non-inlined wrapper functions to inlined functions. >>> +static void gen_ne_fop_FF(DisasContext *dc, int rd, int rs, >> >> 'ne' is for no exception? How about noexcp or something? > > no-exception when it's first introduced. Then after patch 11 it would > become no-env. Preferences for the intermediate stage? Nevermind then.