[Libva] [PATCH 7/8] Execute the 10-bit scaling for P010 surface on Gen9+

[Xiang, Haihao]

On Wed, 2016-11-09 at 14:39 -0500, Zhao Yakui wrote:
> Now the 10-bit scaling based on GPU shader is supported on Gen9+. In
> such
> case it will use the 10-bit scaling based on GPU shader instead of
> three-steps
> by using VEBOX(VEBOX->NV12->Scale NV12->VEBOX).

The 2nd step (NV12 to scale NV12) doesn't use VEBOX.

>  Of course when the size is
> not changed, it still falls back to VEBOX.
> 
> 
> Signed-off-by: Zhao Yakui <yakui.zhao at intel.com>
> ---
>  src/gen75_picture_process.c | 50
> +++++++++++++++++++++++++++++++++++++++------
>  1 file changed, 44 insertions(+), 6 deletions(-)
> 
> diff --git a/src/gen75_picture_process.c
> b/src/gen75_picture_process.c
> index 069088a..4f7f794 100644
> --- a/src/gen75_picture_process.c
> +++ b/src/gen75_picture_process.c
> @@ -38,6 +38,7 @@
>  #include "i965_drv_video.h"
>  #include "i965_post_processing.h"
>  #include "gen75_picture_process.h"
> +#include "gen8_post_processing.h"
>  
>  extern struct hw_context *
>  i965_proc_context_init(VADriverContextP ctx,
> @@ -91,6 +92,21 @@ gen75_vpp_vebox(VADriverContextP ctx,
>       return va_status;
>  } 
>  
> +static int intel_gpe_support_10bit_scaling(struct
> intel_video_process_context *proc_ctx)
> +{
> +    struct i965_proc_context *gpe_proc_ctx;
> +
> +    if (!proc_ctx || !proc_ctx->vpp_fmt_cvt_ctx)
> +        return 0;
> +
> +    gpe_proc_ctx = (struct i965_proc_context *)proc_ctx-
> >vpp_fmt_cvt_ctx;
> +
> +    if (gpe_proc_ctx->pp_context.scaling_context_initialized)
> +        return 1;
> +    else
> +        return 0;
> +}
> +
>  VAStatus 
>  gen75_proc_picture(VADriverContextP ctx,
>                     VAProfile profile,
> @@ -165,12 +181,6 @@ gen75_proc_picture(VADriverContextP ctx,
>          i965_check_alloc_surface_bo(ctx, obj_dst_surf, is_tiled,
> fourcc, sampling);
>      }  
>  
> -    proc_ctx->surface_render_output_object = obj_dst_surf;
> -    proc_ctx->surface_pipeline_input_object = obj_src_surf;
> -    assert(pipeline_param->num_filters <= 4);
> -
> -    int vpp_stage1 = 0, vpp_stage2 = 1, vpp_stage3 = 0;
> -
>      if (pipeline_param->surface_region) {
>          src_rect.x = pipeline_param->surface_region->x;
>          src_rect.y = pipeline_param->surface_region->y;
> @@ -195,6 +205,34 @@ gen75_proc_picture(VADriverContextP ctx,
>          dst_rect.height = obj_dst_surf->orig_height;
>      }
>  
> +    if (pipeline_param->num_filters == 0 || pipeline_param->filters
> == NULL ){
> +        if ((obj_src_surf->fourcc == VA_FOURCC_P010) &&
> +            (obj_dst_surf->fourcc == VA_FOURCC_P010) &&
> +            (src_rect.width != dst_rect.width ||
> +                 src_rect.height != dst_rect.height) &&
> +            intel_gpe_support_10bit_scaling(proc_ctx)) {
> +            struct i965_proc_context *gpe_proc_ctx;
> +            struct i965_surface src_surface, dst_surface;
> +
> +            src_surface.base = (struct object_base *)obj_src_surf;
> +            src_surface.type = I965_SURFACE_TYPE_SURFACE;
> +            dst_surface.base = (struct object_base *)obj_dst_surf;
> +            dst_surface.type = I965_SURFACE_TYPE_SURFACE;
> +            gpe_proc_ctx = (struct i965_proc_context *)proc_ctx-
> >vpp_fmt_cvt_ctx;
> +
> +            return gen9_p010_scaling_post_processing(ctx,
> &gpe_proc_ctx->pp_context,
> +                                                     &src_surface,
> &src_rect,
> +                                                     &dst_surface,
> &dst_rect);
> +        }
> +    }
> +
> +    proc_ctx->surface_render_output_object = obj_dst_surf;
> +    proc_ctx->surface_pipeline_input_object = obj_src_surf;
> +    assert(pipeline_param->num_filters <= 4);
> +
> +    int vpp_stage1 = 0, vpp_stage2 = 1, vpp_stage3 = 0;
> +
> +
>      if(obj_src_surf->fourcc == VA_FOURCC_P010) {
>          vpp_stage1 = 1;
>          vpp_stage2 = 0;