On 2012-10-10 11:54:11 +0100, Måns Rullgård wrote: > Janne Grunau <[email protected]> writes: > > > -void ff_rv40dsp_init_neon(RV34DSPContext *c, DSPContext* dsp) > > +static void ff_rv40dsp_init_neon(RV34DSPContext *c) > > { > > c->put_pixels_tab[0][ 1] = ff_put_rv40_qpel16_mc10_neon; > > c->put_pixels_tab[0][ 3] = ff_put_rv40_qpel16_mc30_neon; > > @@ -136,3 +137,11 @@ void ff_rv40dsp_init_neon(RV34DSPContext *c, > > DSPContext* dsp) > > c->rv40_weak_loop_filter[0] = ff_rv40_h_weak_loop_filter_neon; > > c->rv40_weak_loop_filter[1] = ff_rv40_v_weak_loop_filter_neon; > > } > > + > > +void ff_rv40dsp_init_arm(RV34DSPContext *c, DSPContext* dsp) > > +{ > > + int cpu_flags = av_get_cpu_flags(); > > + > > + if (have_neon(cpu_flags)) > > + ff_rv40dsp_init_neon(c); > > +} > > This will either break or generate warnings with neon disabled.
neither but it depends on dead code elimination which I assumed to be ok when I wrote that. I'll wrap #if HAVE_NEON around it. btw: a couple of objects do not build with --disable-optimizations Janne _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
