Quoting Michel Thierry (2018-02-07 22:34:00)
> On 2/7/2018 1:05 PM, Chris Wilson wrote:
> > +static bool needs_preempt_context(struct drm_i915_private *i915)
> > +{
> > + return HAS_LOGICAL_RING_PREEMPTION(i915);
> > +}
> > +
>
> Wasn't adding a function just for this an overkill?
Maybe over-engineering. Based on the discussion that we may not need a
preempt context for icl (I'm still skeptical), I thought having a
convenient slot to hook in the extra logic was worth the seeming
tautology.
> I agree it makes gem_contexts_init self-explanatory but there are still
> some HAS_LOGICAL_RING_PREEMPTION in guc_init/fini_wq and gvt/scheduler.
A few ordering problems^W challenges prevented killing those remaining
HAS_LOGICAL_RING_PREEMPTION; afaict guc_init is making some decisions
before the contexts have been initialised.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
