[Intel-gfx] [PATCH 1/4] drm/i915: Replace global_seqno with a hangcheck heartbeat seqno

[Chris Wilson]

Quoting Tvrtko Ursulin (2019-02-26 07:34:37)
> 
> On 25/02/2019 18:40, Chris Wilson wrote:
> > Quoting Tvrtko Ursulin (2019-02-25 17:59:40)
> >>
> >> On 25/02/2019 16:23, Chris Wilson wrote:
> >>>    static inline struct i915_priolist *to_priolist(struct rb_node *rb)
> >>>    {
> >>>        return rb_entry(rb, struct i915_priolist, node);
> >>> @@ -2206,6 +2212,10 @@ static u32 *gen8_emit_fini_breadcrumb(struct i915_request *request, u32 *cs)
> >>>                                  request->fence.seqno,
> >>>                                  request->timeline->hwsp_offset);
> >>>    
> >>> +     cs = gen8_emit_ggtt_write(cs,
> >>> +                               intel_engine_next_hangcheck_seqno(request->engine),
> >>> +                               intel_hws_hangcheck_address(request->engine));
> >>> +
> >>>        cs = gen8_emit_ggtt_write(cs,
> >>>                                  request->global_seqno,
> >>>                                  intel_hws_seqno_address(request->engine));
> >>> @@ -2230,6 +2240,11 @@ static u32 *gen8_emit_fini_breadcrumb_rcs(struct i915_request *request, u32 *cs)
> >>>                                      PIPE_CONTROL_FLUSH_ENABLE |
> >>>                                      PIPE_CONTROL_CS_STALL);
> >>>    
> >>> +     cs = gen8_emit_ggtt_write_rcs(cs,
> >>> +                                   intel_engine_next_hangcheck_seqno(request->engine),
> >>> +                                   intel_hws_hangcheck_address(request->engine),
> >>> +                                   PIPE_CONTROL_CS_STALL);
> >>
> >> Are CS_STALL needed on two writes or only last one would be enough? Or
> >> even, should all flushes be moved to the last pipe control?
> > 
> > The CS_STALL is overkill as there's no requirement for it to be before
> > the global_seqno, but the convenience and ease to reason over win.

[snip]

> Ok have at it.
> 
> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin at intel.com>

I was just about to resend without the CS_STALL...
-Chris