[Mesa-dev] [PATCH 20/22] intel/isl/gen4: Represent cube maps with 3D layout

[Jason Ekstrand]

I was sure I'd written this patch at least twice but I can't find it... :-(

On Tue, Jul 18, 2017 at 1:46 AM, Topi Pohjolainen <
topi.pohjolainen at gmail.com> wrote:

> Signed-off-by: Topi Pohjolainen <topi.pohjolainen at intel.com>
> ---
>  src/intel/isl/isl.c | 40 +++++++++++++++++++++++++++++++++++-----
>  1 file changed, 35 insertions(+), 5 deletions(-)
>
> diff --git a/src/intel/isl/isl.c b/src/intel/isl/isl.c
> index 12ffe3bb51..90b36c33bc 100644
> --- a/src/intel/isl/isl.c
> +++ b/src/intel/isl/isl.c
> @@ -548,7 +548,8 @@ isl_choose_image_alignment_el(const struct isl_device
> *dev,
>  static enum isl_dim_layout
>  isl_surf_choose_dim_layout(const struct isl_device *dev,
>                             enum isl_surf_dim logical_dim,
> -                           enum isl_tiling tiling)
> +                           enum isl_tiling tiling,
> +                           isl_surf_usage_flags_t usage)
>  {
>     /* Sandy bridge needs a special layout for HiZ and stencil. */
>     if (ISL_DEV_GEN(dev) == 6 &&
> @@ -584,6 +585,16 @@ isl_surf_choose_dim_layout(const struct isl_device
> *dev,
>        switch (logical_dim) {
>        case ISL_SURF_DIM_1D:
>        case ISL_SURF_DIM_2D:
> +         /* From the G45 PRM Vol. 1a, "6.17.4.1 Hardware Cube Map Layout":
> +          *
> +          * The cube face textures are stored in the same way as 3D
> surfaces
> +          * are stored (see section 6.17.5 for details).  For cube
> surfaces,
> +          * however, the depth is equal to the number of faces (always 6)
> and
> +          * is not reduced for each MIP.
> +          */
> +         if (ISL_DEV_GEN(dev) == 4 && (usage & ISL_SURF_USAGE_CUBE_BIT))
> +            return ISL_DIM_LAYOUT_GEN4_3D;
> +
>           return ISL_DIM_LAYOUT_GEN4_2D;
>        case ISL_SURF_DIM_3D:
>           return ISL_DIM_LAYOUT_GEN4_3D;
> @@ -635,8 +646,11 @@ isl_calc_phys_level0_extent_sa(const struct
> isl_device *dev,
>        break;
>
>     case ISL_SURF_DIM_2D:
> -      assert(dim_layout == ISL_DIM_LAYOUT_GEN4_2D ||
> -             dim_layout == ISL_DIM_LAYOUT_GEN6_STENCIL_HIZ);
> +      if (ISL_DEV_GEN(dev) == 4 && (info->usage &
> ISL_SURF_USAGE_CUBE_BIT))
> +         assert(dim_layout == ISL_DIM_LAYOUT_GEN4_3D);
> +      else
> +         assert(dim_layout == ISL_DIM_LAYOUT_GEN4_2D ||
> +                dim_layout == ISL_DIM_LAYOUT_GEN6_STENCIL_HIZ);
>
>        if (tiling == ISL_TILING_Ys && info->samples > 1)
>           isl_finishme("%s:%s: multisample TileYs layout", __FILE__,
> __func__);
> @@ -952,7 +966,11 @@ isl_calc_phys_total_extent_el_gen4_3d(
>     const struct isl_format_layout *fmtl = isl_format_get_layout(info->
> format);
>
>     assert(info->samples == 1);
> -   assert(phys_level0_sa->array_len == 1);
> +
> +   if (info->usage & ISL_SURF_USAGE_CUBE_BIT)
> +      assert(phys_level0_sa->array_len == 6);
> +   else
> +      assert(phys_level0_sa->array_len == 1);
>

How about:

if (info->dim != ISL_SURF_DIM_3D) {
   /* PRM quote goes here */
   assert(ISL_DEV_GEN(dev) == 4);
   assert(info->usage & ISL_SURF_USAGE_CUBE);
   assert(phys_level0_sa->array_len == 6);
} else {
   assert(phys_level0_sa->array_len == 1);
}


>
>     uint32_t total_w = 0;
>     uint32_t total_h = 0;
> @@ -966,6 +984,18 @@ isl_calc_phys_total_extent_el_gen4_3d(
>        uint32_t level_h = isl_align_npot(isl_minify(H0, l),
> image_align_sa->h);
>        uint32_t level_d = isl_align_npot(isl_minify(D0, l),
> image_align_sa->d);
>
> +      /* From the G45 PRM Vol. 1a, "6.17.4.1 Hardware Cube Map Layout":
> +       *
> +       * The cube face textures are stored in the same way as 3D surfaces
> +       * are stored (see section 6.17.5 for details).  For cube surfaces,
> +       * however, the depth is equal to the number of faces (always 6) and
> +       * is not reduced for each MIP.
> +       */
> +      if (info->usage & ISL_SURF_USAGE_CUBE_BIT) {
> +         assert(ISL_DEV_GEN(dev) == 4);
> +         level_d = 6;
> +      }
>

I don't really like using CUBE_BIT here.  I would much rather we do
something like

level_d = info->dim == ISL_SURF_DIM_3D ? isl_minify(D0, l) : A0;

where A0 is declared outside the loop as the array length which must be 6.
The align isn't really needed for depth as DALIGN isn't a thing.


> +
>        uint32_t max_layers_horiz = MIN(level_d, 1u << l);
>        uint32_t max_layers_vert = isl_align(level_d, 1u << l) / (1u << l);
>
> @@ -1427,7 +1457,7 @@ isl_surf_init_s(const struct isl_device *dev,
>     isl_tiling_get_info(tiling, fmtl->bpb, &tile_info);
>
>     const enum isl_dim_layout dim_layout =
> -      isl_surf_choose_dim_layout(dev, info->dim, tiling);
> +      isl_surf_choose_dim_layout(dev, info->dim, tiling, info->usage);
>
>     enum isl_msaa_layout msaa_layout;
>     if (!isl_choose_msaa_layout(dev, info, tiling, &msaa_layout))
> --
> 2.11.0
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/mesa-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20170718/b64b1184/attachment-0001.html>