[Mesa-dev] [PATCH v2 06/12] i965/gen7+: Implement fast color clear operation in BLORP.

[Eric Anholt]

Paul Berry <stereotype441 at gmail.com> writes:

> +/**
> + * Determine if fast color clear supports the given clear color.
> + *
> + * Fast color clear can only clear to color values of 1.0 or 0.0.  At the
> + * moment we only support floating point buffers.

floating point, unorm, or snorm buffers, right?

> diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
> index a75ac81..ba941c0 100644
> --- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
> +++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
> @@ -1201,6 +1201,54 @@ intel_miptree_alloc_mcs(struct intel_context *intel,
>  #endif
>  }
>  
> +
> +bool
> +intel_miptree_alloc_non_msrt_mcs(struct intel_context *intel,
> +                                 struct intel_mipmap_tree *mt)
> +{
> +#ifdef I915
> +   assert(!"MCS not supported on i915");
> +   return false;
> +#else
> +   assert(mt->mcs_mt == NULL);
> +
> +   /* The format of the MCS buffer is opaque to the driver; all that matters
> +    * is that we get its size and pitch right.  We'll pretend that the format
> +    * is R32.  Since an MCS tile covers 128 blocks horizontally, and a Y-tiled
> +    * R32 buffer is 32 pixels across, we'll need to scale the width down by
> +    * the block width and then a further factor of 4.  Since an MCS tile
> +    * covers 256 blocks vertically, and a Y-tiled R32 buffer is 32 rows high,
> +    * we'll need to scale the height down by the block height and then a
> +    * further factor of 8.
> +    */
> +   const gl_format format = MESA_FORMAT_R_UINT32;
> +   unsigned block_width_px;
> +   unsigned block_height;
> +   intel_get_non_msrt_mcs_alignment(intel, mt, &block_width_px, &block_height);
> +   unsigned width_divisor = block_width_px * 4;
> +   unsigned height_divisor = block_height * 8;
> +   unsigned mcs_width =
> +      ALIGN(mt->logical_width0, width_divisor) / width_divisor;
> +   unsigned mcs_height =
> +      ALIGN(mt->logical_height0, height_divisor) / height_divisor;
> +   assert(mt->logical_depth0 == 1);
> +   mt->mcs_mt = intel_miptree_create(intel,
> +                                     mt->target,
> +                                     format,
> +                                     mt->first_level,
> +                                     mt->last_level,
> +                                     mcs_width,
> +                                     mcs_height,
> +                                     mt->logical_depth0,
> +                                     true,
> +                                     0 /* num_samples */,
> +                                     INTEL_MIPTREE_TILING_Y);
> +
> +   return mt->mcs_mt;
> +#endif

Does this actually create an appropriate miptree?  Don't you need
something that has the smaller levels' sizes also aligned to
width_divisor and height_divisor?
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: not available
URL: <http://lists.freedesktop.org/archives/mesa-dev/attachments/20130611/c9bc3b40/attachment.pgp>