On Wed, Mar 14, 2018 at 09:31:32AM -0700, Lucas De Marchi wrote: > Reorder fields so we save 8 bytes per instance: this removes a 4-bytes > hole after enum intel_dpll_id and a 4-bytes padding. > > Signed-off-by: Lucas De Marchi <[email protected]> > --- > > Is this something desirable? I happened to be looking at > intel_shared_dpll and noticed the hole. I haven't checked any other struct > yet, but there are probably more and more important ones. This one saves > only 8 * I915_NUM_PLLS. > > drivers/gpu/drm/i915/intel_dpll_mgr.h | 10 +++++----- > 1 file changed, 5 insertions(+), 5 deletions(-) > > diff --git a/drivers/gpu/drm/i915/intel_dpll_mgr.h > b/drivers/gpu/drm/i915/intel_dpll_mgr.h > index f24ccf443d25..9635522dcb32 100644 > --- a/drivers/gpu/drm/i915/intel_dpll_mgr.h > +++ b/drivers/gpu/drm/i915/intel_dpll_mgr.h > @@ -238,11 +238,6 @@ struct intel_shared_dpll { > */ > enum intel_dpll_id id; > > - /** > - * @funcs: platform specific hooks > - */ > - struct intel_shared_dpll_funcs funcs; > - > #define INTEL_DPLL_ALWAYS_ON (1 << 0) > /** > * @flags: > @@ -252,6 +247,11 @@ struct intel_shared_dpll { > * not in use by any CRTC. > */ > uint32_t flags; > + > + /** > + * @funcs: platform specific hooks > + */ > + struct intel_shared_dpll_funcs funcs;
Why do we need to copy the entire thing here anyway? Can't we just make this a pointer? > }; > > #define SKL_DPLL0 0 > -- > 2.14.3 > > _______________________________________________ > Intel-gfx mailing list > [email protected] > https://lists.freedesktop.org/mailman/listinfo/intel-gfx -- Ville Syrjälä Intel OTC _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
