Re: [PATCH 1/2] drm/i915: Eliminate unrequired dwords for MI_SET_CONTEXT

[Daniel Vetter <daniel@ffwll.ch>]

On Thu, Aug 08, 2013 at 08:00:25PM +0100, Chris Wilson wrote:
> A later patch adds yet another workaround for MI_SET_CONTEXT, at which
> point we start to end up with more NOOPs than actual command dwords
> along the non-workaround paths. It is time that we made the MI_SET_CONTEXT
> a variable length block and only emit the dwords we truly need.
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Reviewed-by: Ben Widawsky <ben@bwidawsk.net>

Since this seems to be based on top of gen5 context support: Care to slap
an r-b and tested-by onto it so that I can merge the entire shebang?

Cheers, Daniel
> ---
>  drivers/gpu/drm/i915/i915_gem_context.c | 17 +++++++++--------
>  1 file changed, 9 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
> index 879bfa2..8a7b61e 100644
> --- a/drivers/gpu/drm/i915/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/i915_gem_context.c
> @@ -348,7 +348,7 @@ mi_set_context(struct intel_ring_buffer *ring,
>  	       struct i915_hw_context *new_context,
>  	       u32 hw_flags)
>  {
> -	int ret;
> +	int ret, len;
>  
>  	/* w/a: If Flush TLB Invalidation Mode is enabled, driver must do a TLB
>  	 * invalidation prior to MI_SET_CONTEXT. On GEN6 we don't set the value
> @@ -361,7 +361,14 @@ mi_set_context(struct intel_ring_buffer *ring,
>  			return ret;
>  	}
>  
> -	ret = intel_ring_begin(ring, 6);
> +	len = 4;
> +	switch (INTEL_INFO(ring->dev)->gen) {
> +	case 7:
> +	case 5: len += 2;
> +		break;
> +	}
> +
> +	ret = intel_ring_begin(ring, len);
>  	if (ret)
>  		return ret;
>  
> @@ -373,9 +380,6 @@ mi_set_context(struct intel_ring_buffer *ring,
>  	case 5:
>  		intel_ring_emit(ring, MI_SUSPEND_FLUSH | MI_SUSPEND_FLUSH_EN);
>  		break;
> -	default:
> -		intel_ring_emit(ring, MI_NOOP);
> -		break;
>  	}
>  
>  	intel_ring_emit(ring, MI_NOOP);
> @@ -395,9 +399,6 @@ mi_set_context(struct intel_ring_buffer *ring,
>  	case 5:
>  		intel_ring_emit(ring, MI_SUSPEND_FLUSH);
>  		break;
> -	default:
> -		intel_ring_emit(ring, MI_NOOP);
> -		break;
>  	}
>  
>  	intel_ring_advance(ring);
> -- 
> 1.8.4.rc1
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch