[PATCH i-g-t v2 18/18] lib/intel_aux_pgtable: Pick a more optimal aux format for 10bpc

[Juha-Pekka Heikkila]

Reviewed-by: Juha-Pekka Heikkila <juhapekka.heikkila at gmail.com>

On 18.9.2024 15.05, Ville Syrjala wrote:
> From: Ville Syrjälä <ville.syrjala at linux.intel.com>
> 
> Pick the (presumably) more optimal 0x18 aux format for
> 10bpc formats. This matches what Mesa uses as well.
> 
> Signed-off-by: Ville Syrjälä <ville.syrjala at linux.intel.com>
> ---
>   lib/intel_aux_pgtable.c | 6 +++++-
>   1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/lib/intel_aux_pgtable.c b/lib/intel_aux_pgtable.c
> index 8ff48641619e..de0740e950f8 100644
> --- a/lib/intel_aux_pgtable.c
> +++ b/lib/intel_aux_pgtable.c
> @@ -23,6 +23,7 @@
>   #define AUX_FORMAT_ARGB_8B	0x0A
>   #define AUX_FORMAT_NV12_21	0x0F
>   #define AUX_FORMAT_RGBA16_FLOAT	0x10
> +#define AUX_FORMAT_ARGB_10B	0x18
>   
>   struct pgtable_level_desc {
>   	int idx_shift;
> @@ -304,7 +305,10 @@ static uint64_t pgt_get_l1_flags(const struct intel_buf *buf, int surface_idx)
>   	} else {
>   		switch (buf->bpp) {
>   		case 32:
> -			entry.e.format = AUX_FORMAT_ARGB_8B;
> +			if (buf->depth == 30)
> +				entry.e.format = AUX_FORMAT_ARGB_10B;
> +			else
> +				entry.e.format = AUX_FORMAT_ARGB_8B;
>   			entry.e.depth = bpp_to_depth_val(32);
>   			break;
>   		case 64: