On Tue, 2017-09-05 at 13:17 -0700, Oscar Mateo wrote:
>
> On 08/29/2017 04:07 PM, Rodrigo Vivi wrote:
> > WA to enable HW L1 Banking fix that allows aniso to operate
> > at full sample rate.
> >
> > Cc: Mika Kuoppala <mika.kuoppala@xxxxxxxxx>
> > Cc: Oscar Mateo <oscar.mateo@xxxxxxxxx>
> > Cc: Ben Widawsky <ben@xxxxxxxxxxxx>
> > Cc: Anuj Phogat <anuj.phogat@xxxxxxxxx>
> > Signed-off-by: Rodrigo Vivi <rodrigo.vivi@xxxxxxxxx>
> > ---
> > drivers/gpu/drm/i915/i915_reg.h | 1 +
> > drivers/gpu/drm/i915/intel_engine_cs.c | 3 +++
> > 2 files changed, 4 insertions(+)
> >
> > diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> > index e2908ae34004..1ad22a824921 100644
> > --- a/drivers/gpu/drm/i915/i915_reg.h
> > +++ b/drivers/gpu/drm/i915/i915_reg.h
> > @@ -8072,6 +8072,7 @@ enum {
> > #define HSW_SAMPLE_C_PERFORMANCE (1<<9)
> > #define GEN8_CENTROID_PIXEL_OPT_DIS (1<<8)
> > #define GEN9_DISABLE_OCL_OOB_SUPPRESS_LOGIC (1<<5)
> > +#define CNL_FAST_ANISO_L1_BANKING_FIX (1<<4)
>
> Why CNL_ instead of GEN10_ prefix? (out of curiosity)
hmm... I guess I should have used gen10 for compatibility with other
bits around this... :/
But I pick CNL instead because it is easier to match the spec that
has bit started for CNL as:
GEN:BUG:1937670 [CNL...
if it had started with GEN10:etc I would have picked the GEN10 I
guess...
>
> > #define GEN8_SAMPLER_POWER_BYPASS_DIS (1<<1)
> >
> > #define GEN9_HALF_SLICE_CHICKEN7 _MMIO(0xe194)
> > diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c
> > index a6ac9d0a4156..4b9b7828802d 100644
> > --- a/drivers/gpu/drm/i915/intel_engine_cs.c
> > +++ b/drivers/gpu/drm/i915/intel_engine_cs.c
> > @@ -1090,6 +1090,9 @@ static int cnl_init_workarounds(struct intel_engine_cs *engine)
> > /* WaPushConstantDereferenceHoldDisable:cnl */
> > WA_SET_BIT(GEN7_ROW_CHICKEN2, PUSH_CONSTANT_DEREF_DISABLE);
> >
> > + /* FtrEnableFastAnisoL1BankingFix: cnl */
> > + WA_SET_BIT_MASKED(HALF_SLICE_CHICKEN3, CNL_FAST_ANISO_L1_BANKING_FIX);
> > +
>
> I have this only for B0 onwards, but I guess you glossed over this
> because you know more than me about which CNL stepping is the production
> one, so:
yeap, we don't have A0 available anyways...
And with no A0 related wa implemented, so our support is B0+...
> Rviewed-by: Oscar Mateo <oscar.mateo@xxxxxxxxx>
Thanks, although it is merged already...
>
> > /* WaEnablePreemptionGranularityControlByUMD:cnl */
> > ret= wa_ring_whitelist_reg(engine, GEN8_CS_CHICKEN1);
> > if (ret)
>
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
