[Intel-gfx] [PATCH 11/18] drm/i915: Record space required for request emission

[Tvrtko Ursulin]

On 14/09/2016 07:52, Chris Wilson wrote:
> In the next patch, we will use deferred request emission. That requires
> reserving sufficient space in the ringbuffer to emit the request, which
> first requires us to know how large the request is.
>
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> ---
>   drivers/gpu/drm/i915/i915_gem_request.c |  1 +
>   drivers/gpu/drm/i915/intel_lrc.c        |  6 ++++++
>   drivers/gpu/drm/i915/intel_ringbuffer.c | 29 +++++++++++++++++++++++++++--
>   drivers/gpu/drm/i915/intel_ringbuffer.h |  1 +
>   4 files changed, 35 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
> index ebc2feba3e50..9a735e2d5aea 100644
> --- a/drivers/gpu/drm/i915/i915_gem_request.c
> +++ b/drivers/gpu/drm/i915/i915_gem_request.c
> @@ -425,6 +425,7 @@ i915_gem_request_alloc(struct intel_engine_cs *engine,
>   	 * away, e.g. because a GPU scheduler has deferred it.
>   	 */
>   	req->reserved_space = MIN_SPACE_FOR_ADD_REQUEST;
> +	GEM_BUG_ON(req->reserved_space < engine->emit_request_sz);
>   
>   	if (i915.enable_execlists)
>   		ret = intel_logical_ring_alloc_request_extras(req);
> diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
> index 00fcf36ba919..7e944a0acc10 100644
> --- a/drivers/gpu/drm/i915/intel_lrc.c
> +++ b/drivers/gpu/drm/i915/intel_lrc.c
> @@ -1572,6 +1572,8 @@ static int gen8_emit_request(struct drm_i915_gem_request *request)
>   	return intel_logical_ring_advance(request);
>   }
>   
> +static const int gen8_emit_request_sz = 6 + WA_TAIL_DWORDS;
> +
>   static int gen8_emit_request_render(struct drm_i915_gem_request *request)
>   {
>   	struct intel_ring *ring = request->ring;
> @@ -1603,6 +1605,8 @@ static int gen8_emit_request_render(struct drm_i915_gem_request *request)
>   	return intel_logical_ring_advance(request);
>   }
>   
> +static const int gen8_emit_request_render_sz = 8 + WA_TAIL_DWORDS;
> +
>   static int gen8_init_rcs_context(struct drm_i915_gem_request *req)
>   {
>   	int ret;
> @@ -1677,6 +1681,7 @@ logical_ring_default_vfuncs(struct intel_engine_cs *engine)
>   	engine->reset_hw = reset_common_ring;
>   	engine->emit_flush = gen8_emit_flush;
>   	engine->emit_request = gen8_emit_request;
> +	engine->emit_request_sz = gen8_emit_request_sz;
>   	engine->submit_request = execlists_submit_request;
>   
>   	engine->irq_enable = gen8_logical_ring_enable_irq;
> @@ -1799,6 +1804,7 @@ int logical_render_ring_init(struct intel_engine_cs *engine)
>   	engine->init_context = gen8_init_rcs_context;
>   	engine->emit_flush = gen8_emit_flush_render;
>   	engine->emit_request = gen8_emit_request_render;
> +	engine->emit_request_sz = gen8_emit_request_render_sz;
>   
>   	ret = intel_engine_create_scratch(engine, 4096);
>   	if (ret)
> diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
> index 597e35c9b699..c8c9ad40fd93 100644
> --- a/drivers/gpu/drm/i915/intel_ringbuffer.c
> +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
> @@ -1405,6 +1405,8 @@ static int i9xx_emit_request(struct drm_i915_gem_request *req)
>   	return 0;
>   }
>   
> +static const int i9xx_emit_request_sz = 4;
> +
>   /**
>    * gen6_sema_emit_request - Update the semaphore mailbox registers
>    *
> @@ -1458,6 +1460,8 @@ static int gen8_render_emit_request(struct drm_i915_gem_request *req)
>   	return 0;
>   }
>   
> +static const int gen8_render_emit_request_sz = 8;
> +
>   /**
>    * intel_ring_sync - sync the waiter to the signaller on seqno
>    *
> @@ -2677,8 +2681,21 @@ static void intel_ring_default_vfuncs(struct drm_i915_private *dev_priv,
>   	engine->reset_hw = reset_ring_common;
>   
>   	engine->emit_request = i9xx_emit_request;
> -	if (i915.semaphores)
> +	engine->emit_request_sz = i9xx_emit_request_sz;
> +	if (i915.semaphores) {
> +		int num_rings;
> +
>   		engine->emit_request = gen6_sema_emit_request;
> +
> +		num_rings = hweight32(INTEL_INFO(dev_priv)->ring_mask) - 1;

You can use INTEL_INFO(dev_priv)->num_rings instead of hweight32.

> +		if (INTEL_GEN(dev_priv) >= 8) {
> +			engine->emit_request_sz += num_rings * 6;
> +		} else {
> +			engine->emit_request_sz += num_rings * 3;
> +			if (num_rings & 1)
> +				engine->emit_request_sz++;
> +		}
> +	}
>   	engine->submit_request = i9xx_submit_request;
>   
>   	if (INTEL_GEN(dev_priv) >= 8)
> @@ -2706,9 +2723,17 @@ int intel_init_render_ring_buffer(struct intel_engine_cs *engine)
>   	if (INTEL_GEN(dev_priv) >= 8) {
>   		engine->init_context = intel_rcs_ctx_init;
>   		engine->emit_request = gen8_render_emit_request;
> +		engine->emit_request_sz = gen8_render_emit_request_sz;
>   		engine->emit_flush = gen8_render_ring_flush;
> -		if (i915.semaphores)
> +		if (i915.semaphores) {
> +			int num_rings;
> +
>   			engine->semaphore.signal = gen8_rcs_signal;
> +
> +			num_rings =
> +				hweight32(INTEL_INFO(dev_priv)->ring_mask) - 1;

Here as well.

> +			engine->emit_request_sz += num_rings * 6;
> +		}
>   	} else if (INTEL_GEN(dev_priv) >= 6) {
>   		engine->init_context = intel_rcs_ctx_init;
>   		engine->emit_flush = gen7_render_ring_flush;
> diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
> index d6f80b3d66b4..47ad19f5441d 100644
> --- a/drivers/gpu/drm/i915/intel_ringbuffer.h
> +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
> @@ -226,6 +226,7 @@ struct intel_engine_cs {
>   #define I915_DISPATCH_PINNED BIT(1)
>   #define I915_DISPATCH_RS     BIT(2)
>   	int		(*emit_request)(struct drm_i915_gem_request *req);
> +	int		emit_request_sz;
>   
>   	/* Pass the request to the hardware queue (e.g. directly into
>   	 * the legacy ringbuffer or to the end of an execlist).

Regards,

Tvrtko