| Message ID | 1466418211-3459-1-git-send-email-david.s.gordon@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Mon, Jun 20, 2016 at 11:23:31AM +0100, Dave Gordon wrote: > If a context waiting for VBlank were switched out, the GuC would > have to receive the VBlank interrupt so that it could resubmit > the context. However, we don't use the GuC internal scheduler, > and we always set the CTX_CTRL_INHIBIT_SYN_CTX_SWITCH bit in the > RING_CONTEXT_CONTROL register, so this case cannot arise. > > Consequently, the GuC doesn't need to see VBlanks, and we may be > waking it up unnecessarily by sending them. So let's not ... > > Signed-off-by: Dave Gordon <david.s.gordon@intel.com> Missing v2, and commit message should mention why we shut vblanks down entirely for the guc. With that fixed: Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> I'm soonish on vacations, so pls ask Tvrkto or someone else to push. -Daniel > --- > drivers/gpu/drm/i915/intel_guc_loader.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/drivers/gpu/drm/i915/intel_guc_loader.c b/drivers/gpu/drm/i915/intel_guc_loader.c > index 8fe96a2..5d989d2 100644 > --- a/drivers/gpu/drm/i915/intel_guc_loader.c > +++ b/drivers/gpu/drm/i915/intel_guc_loader.c > @@ -105,8 +105,8 @@ static void direct_interrupts_to_guc(struct drm_i915_private *dev_priv) > int irqs; > u32 tmp; > > - /* tell all command streamers to forward interrupts and vblank to GuC */ > - irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_ALWAYS); > + /* tell all command streamers to forward interrupts (but not vblank) to GuC */ > + irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_NEVER); > irqs |= _MASKED_BIT_ENABLE(GFX_INTERRUPT_STEERING); > for_each_engine(engine, dev_priv) > I915_WRITE(RING_MODE_GEN7(engine), irqs); > -- > 1.9.1 >
diff --git a/drivers/gpu/drm/i915/intel_guc_loader.c b/drivers/gpu/drm/i915/intel_guc_loader.c index 8fe96a2..5d989d2 100644 --- a/drivers/gpu/drm/i915/intel_guc_loader.c +++ b/drivers/gpu/drm/i915/intel_guc_loader.c @@ -105,8 +105,8 @@ static void direct_interrupts_to_guc(struct drm_i915_private *dev_priv) int irqs; u32 tmp; - /* tell all command streamers to forward interrupts and vblank to GuC */ - irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_ALWAYS); + /* tell all command streamers to forward interrupts (but not vblank) to GuC */ + irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_NEVER); irqs |= _MASKED_BIT_ENABLE(GFX_INTERRUPT_STEERING); for_each_engine(engine, dev_priv) I915_WRITE(RING_MODE_GEN7(engine), irqs);
If a context waiting for VBlank were switched out, the GuC would have to receive the VBlank interrupt so that it could resubmit the context. However, we don't use the GuC internal scheduler, and we always set the CTX_CTRL_INHIBIT_SYN_CTX_SWITCH bit in the RING_CONTEXT_CONTROL register, so this case cannot arise. Consequently, the GuC doesn't need to see VBlanks, and we may be waking it up unnecessarily by sending them. So let's not ... Signed-off-by: Dave Gordon <david.s.gordon@intel.com> --- drivers/gpu/drm/i915/intel_guc_loader.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)