On Wed, Sep 04, 2013 at 08:50:13PM +0200, Daniel Vetter wrote: > On Fri, Aug 30, 2013 at 1:50 AM, <mengdong.lin@xxxxxxxxx> wrote: > > + /* Wait for 2 vertical blanks */ > > + intel_wait_for_vblank(dev, pipe); > > + intel_wait_for_vblank(dev, pipe); > > + > > + /* Disable audio PD. This is optional as per Bspec. */ > > + temp = I915_READ(HSW_AUD_PIN_ELD_CP_VLD); > > + temp &= ~(AUDIO_OUTPUT_ENABLE_A << (pipe * 4)); > > + I915_WRITE(HSW_AUD_PIN_ELD_CP_VLD, temp); > > If this is optional do we really need the two vblank waits above? > Adding them just for fun when we generally try to rip out as many > vblank waits as possible from the modeset code isn't all that great > ... One idea I had for these kinds of vblank waits (there also one required for IPS for instance) is that we might just sample a vblank counter after the first step, then at the latest point we can, we'd wait for the frame counter to have passed the sampled vaoue + whatever extra is needed. That might allow us to do other stuff in parallel while the required number of vblanks will elapese. -- Ville Syrjälä Intel OTC _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
