NAME

SYNOPSIS

DESCRIPTION

HINT_CURSOR - the layer will be treated as a cursor layer and implies alpha blending. This may be updated out of the normal processing pipeline and as part of a raw input handler where other graphics functions might be unavailable.

There is also an experimental HINT_DIRECT flag that, if set, will disable any platform controlled heuristic between direct scanout and forced composition. This is an optimization that has a number of driver and backend bugs still. For normal/simple display configurations it is likely to work, but with edge cases for multi-display hotplug, rotated secondary displays and so on. This is marked experimental, but if the feature is deprecated that flag will simply resolve to 0 so it is API -wise safe for use. The caveats should be conveyed to a user.

NOTES

1

A src referencing an object with a feed- function, such as one coming from define_recordtarget , define_calctarget and so on, is a terminal state transition.

2

Multiple output paths with varying performance characteristics may apply here, including forced repacking stages that are very costly. Avoid using custom texture coordinate sets, custom shaders or using the

src object as part of other rendering operations to ensure that any fast scanout path can be applied.

3

A rendertarget as src has a different coordinate transform when going from 'as a texture' to 'as an output' behavior and may end up inverted. The engine will attempt to account for this, but only if no custom texture coordinate set has been defined for the object.

4

When calling map_video_display with a rendertarget as src, its projection transform may be rebuilt. This will override any coordinate system changes made by image_resize_storage which then needs to be repeated.

5

Mapping the same rendertarget to multiple displays is undefined behaviour. The reason is that the backing store of src might need to mutate to fit the scanout characteristics of display . Even for the same display, these ted to have slight changes that make sharing semantics have subtle edge cases that are hard to predict. The best workaround for this is to create additional rendertargets through define_linktarget .

EXAMPLE

function map_video_display0()
      resize_video_canvas(640, 480);
      img_1 = fill_surface(320, 200, 255, 0, 0);
      img_2 = fill_surface(320, 200, 0, 255, 0);
      img_3 = fill_surface(320, 200, 0, 0, 255);
      img_4 = fill_surface(320, 200, 255, 0, 255);
      show_image({img_1, img_2, img_3, img_4});
      move_image(img_2, 320, 0);
      move_image(img_3, 0, 200);
      move_image(img_4, 320, 200);
      a = null_surface(640, 480);
      move_image(a, 320, 240, 1000);
      image_sharestorage(WORLDID, a);
      outputs = {};
      list = video_displaymodes();
      uniq = 0;
      for i,v in ipairs(list) do
            if (outputs[v.displayid] == nil) then
                  outputs[v.displayid] = v;
                  uniq = uniq + 1;
            end
      end
      if (uniq < 2) then
            return shutdown("insufficient outputs");
      end
      local first = nil;
      local second = nil;
      for k,v in pairs(list) do
            if (first == nil) then
                  first = k;
            elseif (second == nil and k ~= first) then
                  second = k;
                  break;
            end
      end
      if (second == nil) then
            return shutdown("no secondary display output found.");
      end
      map_video_display(first, WORLDID);
      map_video_display(second, next_a);
end

MISUSE

function map_video_display0()
      map_video_display("random", WORLDID);
end

SEE ALSO: