[Intel-gfx] [PATCH v3] drm/i915/dg2: Add performance workaround 18019455067

Matt Roper matthew.d.roper at intel.com
Mon Jul 25 23:23:24 UTC 2022


I think you may have missed Lucas' reply to your v3:

https://lists.freedesktop.org/archives/intel-gfx/2022-June/300712.html

Also, here's the reply to v2 that he's referring to:

https://lists.freedesktop.org/archives/intel-gfx/2022-June/300646.html

I.e., he wants this to be called from a new 'tuning_init' function that
is itself called from general_render_compute_wa_init, since we expect
more of these things to show up in the future so it makes sense to have
a dedicated place for them.


Matt

On Wed, Jul 20, 2022 at 11:19:18AM +0300, Lionel Landwerlin wrote:
> Ping?
> 
> On 11/07/2022 14:30, Lionel Landwerlin wrote:
> > Ping?
> > 
> > On 30/06/2022 11:35, Lionel Landwerlin wrote:
> > > The recommended number of stackIDs for Ray Tracing subsystem is 512
> > > rather than 2048 (default HW programming).
> > > 
> > > v2: Move the programming to dg2_ctx_gt_tuning_init() (Lucas)
> > > 
> > > v3: Move programming to general_render_compute_wa_init() (Matt)
> > > 
> > > Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin at intel.com>
> > > ---
> > >   drivers/gpu/drm/i915/gt/intel_gt_regs.h     | 4 ++++
> > >   drivers/gpu/drm/i915/gt/intel_workarounds.c | 9 +++++++++
> > >   2 files changed, 13 insertions(+)
> > > 
> > > diff --git a/drivers/gpu/drm/i915/gt/intel_gt_regs.h
> > > b/drivers/gpu/drm/i915/gt/intel_gt_regs.h
> > > index 07ef111947b8c..12fc87b957425 100644
> > > --- a/drivers/gpu/drm/i915/gt/intel_gt_regs.h
> > > +++ b/drivers/gpu/drm/i915/gt/intel_gt_regs.h
> > > @@ -1112,6 +1112,10 @@
> > >   #define   GEN12_PUSH_CONST_DEREF_HOLD_DIS    REG_BIT(8)
> > >     #define RT_CTRL                    _MMIO(0xe530)
> > > +#define   RT_CTRL_NUMBER_OF_STACKIDS_MASK    REG_GENMASK(6, 5)
> > > +#define   NUMBER_OF_STACKIDS_512        2
> > > +#define   NUMBER_OF_STACKIDS_1024        1
> > > +#define   NUMBER_OF_STACKIDS_2048        0
> > >   #define   DIS_NULL_QUERY            REG_BIT(10)
> > >     #define EU_PERF_CNTL1                _MMIO(0xe558)
> > > diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c
> > > b/drivers/gpu/drm/i915/gt/intel_workarounds.c
> > > index 3213c593a55f4..ea674e456cd76 100644
> > > --- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
> > > +++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
> > > @@ -2737,6 +2737,15 @@ general_render_compute_wa_init(struct
> > > intel_engine_cs *engine, struct i915_wa_li
> > >           wa_write_or(wal, VDBX_MOD_CTRL, FORCE_MISS_FTLB);
> > >           wa_write_or(wal, VEBX_MOD_CTRL, FORCE_MISS_FTLB);
> > >       }
> > > +
> > > +    if (IS_DG2(i915)) {
> > > +        /* Performance tuning for Ray-tracing */
> > > +        wa_write_clr_set(wal,
> > > +                 RT_CTRL,
> > > +                 RT_CTRL_NUMBER_OF_STACKIDS_MASK,
> > > + REG_FIELD_PREP(RT_CTRL_NUMBER_OF_STACKIDS_MASK,
> > > +                        NUMBER_OF_STACKIDS_512));
> > > +    }
> > >   }
> > >     static void
> > 
> > 
> 

-- 
Matt Roper
Graphics Software Engineer
VTT-OSGC Platform Enablement
Intel Corporation


More information about the Intel-gfx mailing list