[Intel-gfx] [PATCH 2/3] drm/i915/tgl: Implement Wa_1604555607

Chris Wilson chris at chris-wilson.co.uk
Fri Jul 26 13:20:54 UTC 2019


Quoting Tvrtko Ursulin (2019-07-26 14:15:51)
> 
> On 26/07/2019 01:10, Chris Wilson wrote:
> > Quoting Lucas De Marchi (2019-07-26 01:02:25)
> >> From: Michel Thierry <michel.thierry at intel.com>
> >>
> >> Implement Wa_1604555607 (set the DS pairing timer to 128 cycles).
> >> FF_MODE2 is part of the register state context, that's why it is
> >> implemented here.
> >>
> >> Signed-off-by: Michel Thierry <michel.thierry at intel.com>
> >> Signed-off-by: Lucas De Marchi <lucas.demarchi at intel.com>
> >> ---
> >>   drivers/gpu/drm/i915/gt/intel_workarounds.c | 7 +++++++
> >>   drivers/gpu/drm/i915/i915_reg.h             | 5 +++++
> >>   2 files changed, 12 insertions(+)
> >>
> >> diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c
> >> index a6eb9c6e87ec..3235ef355dfd 100644
> >> --- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
> >> +++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
> >> @@ -572,6 +572,13 @@ static void icl_ctx_workarounds_init(struct intel_engine_cs *engine,
> >>   static void tgl_ctx_workarounds_init(struct intel_engine_cs *engine,
> >>                                       struct i915_wa_list *wal)
> >>   {
> >> +       u32 val;
> >> +
> >> +       /* Wa_1604555607:tgl */
> >> +       val = intel_uncore_read(engine->uncore, FF_MODE2);
> >> +       val &= ~FF_MODE2_TDS_TIMER_MASK;
> >> +       val |= FF_MODE2_TDS_TIMER_128;
> >> +       wa_write_masked_or(wal, FF_MODE2, FF_MODE2_TDS_TIMER_MASK, val);
> > 
> > It will do a rmw on application, so you just need
> >       wa_write_masked_or(wal, FF_MODE2,
> >                          FF_MODE2_TDS_TIMER_MASK, FF_MODE2_TDS_TIMER_128);
> 
> Not with ctx was unfortunately, no rmw there, just lri.

Odd that we read from outside the ctx then, no?

We can do rmw inside ctx_wa if we have to thanks to MI_MATH. Should we
start preparing for that nightmare? :)
-Chris


More information about the Intel-gfx mailing list