On 11 October 2011 22:30, Stefan Dösinger stefan@codeweavers.com wrote:
- if (wined3d_settings.offscreen_rendering_mode == ORM_FBO)
- {
if (!((dst_format->flags & WINED3DFMT_FLAG_FBO_ATTACHABLE) || (dst_usage & WINED3DUSAGE_RENDERTARGET)))return FALSE;- }
- else if (!(dst_usage & WINED3DUSAGE_RENDERTARGET))
- {
return FALSE;- }
This is redundant. Formats are never FBO-attachable if we don't have FBOs. It's not clear this is what you want either way though. This pretty much looks like a copy/paste from fbo_blit_supported(), and the WINED3DUSAGE_RENDERTARGET check there is to allow formats that don't have a FBO-attachable internal format, but were created with the rtInternal format (which should always be attachable). There's really no equivalent check for backbuffer ORM, it just renders something and hopes we can read it back.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Am 12.10.2011 um 11:26 schrieb Henri Verbeet:
On 11 October 2011 22:30, Stefan Dösinger stefan@codeweavers.com wrote:
- if (wined3d_settings.offscreen_rendering_mode == ORM_FBO)
- {
if (!((dst_format->flags & WINED3DFMT_FLAG_FBO_ATTACHABLE) || (dst_usage & WINED3DUSAGE_RENDERTARGET)))return FALSE;- }
- else if (!(dst_usage & WINED3DUSAGE_RENDERTARGET))
- {
return FALSE;- }
Formats are never FBO-attachable if we don't have FBOs. It's not clear this is what you want either way though. This pretty much looks like a copy/paste from fbo_blit_supported(), and the WINED3DUSAGE_RENDERTARGET check there is to allow formats that don't have a FBO-attachable internal format, but were created with the rtInternal format (which should always be attachable). There's really no equivalent check for backbuffer ORM, it just renders something and hopes we can read it back.
I'm not sure I see what you mean. What I want is to make sure the blitter doesn't draw to a surface that GL can't draw to. In the FBO ORM case the fbo_attachable || rendertarget case seems just right, for the reasons you mention.
With backbuffer ORM we don't really know where we can draw(e.g. stuff like surface size vs framebuffer size comes into play), but I think RENDERTARGET usage is a decent heuristic since we'll probably have to be able to draw to it anyways, or things fails.
I could add some surface_is_drawable() function or equivalent surface flag. That might be a good idea on its own, but I don't see a better way to define if a surface can be rendered to than the above statement.
-
Henri Verbeet -
Stefan Dösinger