[Intel-gfx] [PATCH] drm/i915/chv: Add additional workarounds for CHV

[Daniel Vetter]

On Thu, Dec 04, 2014 at 12:58:40PM +0200, Ville Syrjälä wrote:
> On Wed, Dec 03, 2014 at 07:15:31PM +0000, Arun Siluvery wrote:
> > These w/a were recently identified while debugging another issue,
> > +WaClearFlowControlGpgpuContextSave:chv
> > +Wa4x4STCOptimizationDisable:chv
> > 
> > For: VIZ-4090
> > Change-Id: I08d2176dec609396c3a7c2e48b2413e233799fc4
> > Signed-off-by: Arun Siluvery <arun.siluvery at linux.intel.com>
> > ---
> >  drivers/gpu/drm/i915/i915_reg.h         | 1 +
> >  drivers/gpu/drm/i915/intel_ringbuffer.c | 5 +++++
> >  2 files changed, 6 insertions(+)
> > 
> > diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> > index dc03fac..7c7663f 100644
> > --- a/drivers/gpu/drm/i915/i915_reg.h
> > +++ b/drivers/gpu/drm/i915/i915_reg.h
> > @@ -6131,6 +6131,7 @@ enum punit_power_well {
> >  #define   GEN9_DG_MIRROR_FIX_ENABLE	(1<<5)
> >  
> >  #define GEN8_ROW_CHICKEN		0xe4f0
> > +#define   FLOW_CONTROL_ENABLE		(1<<15)
> >  #define   PARTIAL_INSTRUCTION_SHOOTDOWN_DISABLE	(1<<8)
> >  #define   STALL_DOP_GATING_DISABLE		(1<<5)
> >  
> > diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
> > index 79b4ca5..525c9bf 100644
> > --- a/drivers/gpu/drm/i915/intel_ringbuffer.c
> > +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
> > @@ -796,8 +796,10 @@ static int chv_init_workarounds(struct intel_engine_cs *ring)
> >  
> >  	/* WaDisablePartialInstShootdown:chv */
> >  	/* WaDisableThreadStallDopClockGating:chv */
> > +	/* WaClearFlowControlGpgpuContextSave:chv */
> >  	WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN,
> >  			  PARTIAL_INSTRUCTION_SHOOTDOWN_DISABLE |
> > +			  FLOW_CONTROL_ENABLE |
> 
> This seems to be the default value already. I suppose there's no real
> harm in setting it though, but then we should also do the same for BDW.
> 
> Would be nice if we actually did a GPU reset on driver load to guarantee
> a more consistent initial state...

We have some half-baked testcase to compare the different wa registers
states after driver load, reset and suspend/resume. But that's not yet
been fixed up to also take all the static wa settings into account.

Would be awesome if that would finally happen.

/me dreaming ;-)

Cheers, Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch