Re: [Intel-gfx] [PATCH v2 01/14] drm/i915: Keep a global seqno per-engine

2017-02-16 Thread Chris Wilson
On Thu, Feb 16, 2017 at 08:10:07AM +, Tvrtko Ursulin wrote: > > On 15/02/2017 21:49, Chris Wilson wrote: > >On Wed, Feb 15, 2017 at 05:05:40PM +, Tvrtko Ursulin wrote: > >> > >>On 14/02/2017 09:54, Chris Wilson wrote: > >>>Replace the global device seqno with one for each engine, and accou

Re: [Intel-gfx] [PATCH v2 01/14] drm/i915: Keep a global seqno per-engine

2017-02-16 Thread Tvrtko Ursulin
On 15/02/2017 21:49, Chris Wilson wrote: On Wed, Feb 15, 2017 at 05:05:40PM +, Tvrtko Ursulin wrote: On 14/02/2017 09:54, Chris Wilson wrote: Replace the global device seqno with one for each engine, and account for in-flight seqno on each separately. This is consistent with dma-fence as

Re: [Intel-gfx] [PATCH v2 01/14] drm/i915: Keep a global seqno per-engine

2017-02-15 Thread Chris Wilson
On Wed, Feb 15, 2017 at 09:49:41PM +, Chris Wilson wrote: > > >-static int reserve_global_seqno(struct drm_i915_private *i915) > > >+static int reserve_global_seqno(struct intel_engine_cs *engine) > > > > Rename to reserve_engine_seqno? > > Yes. > > > >- u32 active_requests = ++i915->gt.act

Re: [Intel-gfx] [PATCH v2 01/14] drm/i915: Keep a global seqno per-engine

2017-02-15 Thread Chris Wilson
On Wed, Feb 15, 2017 at 09:49:41PM +, Chris Wilson wrote: > > Protected by the timeline spinlock? > > struct_mutex. (Still thinking about what mutex it will be in the > future.) Hmm, was thinking it have already replaced active_requests. However, it is still tied to request allocation/dealloc

Re: [Intel-gfx] [PATCH v2 01/14] drm/i915: Keep a global seqno per-engine

2017-02-15 Thread Chris Wilson
On Wed, Feb 15, 2017 at 05:05:40PM +, Tvrtko Ursulin wrote: > > On 14/02/2017 09:54, Chris Wilson wrote: > >Replace the global device seqno with one for each engine, and account > >for in-flight seqno on each separately. This is consistent with > >dma-fence as each timeline has separate fence-

Re: [Intel-gfx] [PATCH v2 01/14] drm/i915: Keep a global seqno per-engine

2017-02-15 Thread Tvrtko Ursulin
On 14/02/2017 09:54, Chris Wilson wrote: Replace the global device seqno with one for each engine, and account for in-flight seqno on each separately. This is consistent with dma-fence as each timeline has separate fence-contexts for each engine and a seqno is only ordered within a fence-context

[Intel-gfx] [PATCH v2 01/14] drm/i915: Keep a global seqno per-engine

2017-02-14 Thread Chris Wilson
Replace the global device seqno with one for each engine, and account for in-flight seqno on each separately. This is consistent with dma-fence as each timeline has separate fence-contexts for each engine and a seqno is only ordered within a fence-context (i.e. seqno do not need to be ordered wrt