st/va/postproc: use progressive target buffer for scaling

Scaling between interlaced buffers, esp. for scale-up, because
blit will scale up top filed and bottom field separately. it'll
result in the weaving for these buffer with lack of accuracy.
So use shader deint for the case.

Acked-by: Christian König <christian.koenig@amd.com>

diff --git a/src/gallium/state_trackers/va/postproc.c b/src/gallium/state_trackers/va/postproc.c
index 4449126..d3eafe9 100644 (file)
--- a/src/gallium/state_trackers/va/postproc.c
+++ b/src/gallium/state_trackers/va/postproc.c
@@ -118,15 +118,33 @@ static VAStatus vlVaPostProcBlit(vlVaDriver *drv, vlVaContext *context,
    struct pipe_surface **dst_surfaces;
    struct u_rect src_rect;
    struct u_rect dst_rect;
+   bool scale = false;
    unsigned i;
 
    if (src->interlaced != dst->interlaced && dst->interlaced)
       return VA_STATUS_ERROR_INVALID_SURFACE;
 
+   if ((src->width != dst->width || src->height != dst->height) &&
+       (src->interlaced && dst->interlaced))
+      scale = true;
+
    src_surfaces = src->get_surfaces(src);
    if (!src_surfaces || !src_surfaces[0])
       return VA_STATUS_ERROR_INVALID_SURFACE;
 
+   if (scale) {
+      vlVaSurface *surf;
+
+      surf = handle_table_get(drv->htab, context->target_id);
+      surf->templat.interlaced = false;
+      dst->destroy(dst);
+
+      if (vlVaHandleSurfaceAllocate(drv, surf, &surf->templat) != VA_STATUS_SUCCESS)
+         return VA_STATUS_ERROR_ALLOCATION_FAILED;
+
+      dst = context->target = surf->buffer;
+   }
+
    dst_surfaces = dst->get_surfaces(dst);
    if (!dst_surfaces || !dst_surfaces[0])
       return VA_STATUS_ERROR_INVALID_SURFACE;