For each pixel format we store a flag in the table whether it supports
post pixelshader blending. Before applying blending or during a
context switch we verify that blending is turned off for the
format. In case of R32F this gave a 5-6x performance boost (without
filtering and software conversion).
This adds code for handling fixed function fragment processing with the
GL_ATI_fragment_shader extension. This is a sort-of programmable
interface for fragment processing at the level of shader model 1.4 in
d3d. This code is of use on r200, r250 and r280 cards(radeon 8500 to
9200) which do not support GL_ARB_fragment_program, but support pixel
shader 1.4 on Windows. This code is somewhat a counterpart to the
existing fragment processing code using GL_NV_register_combiners and
GL_NV_texture_shader.
Add a new property of the shader backend which indicates whether the
shader backend is able to dirtify single constants rather than
dirtifying vshader and pshader constants as a whole. Depending on this
a different Set*ConstantF implementation is used which marks constants
dirty. The ARB shader backend uses this and marks constants clean
after uploading.
Before it was done in findContext, before selecting the new context
which is bad (it doesn't always work). The new code works and this
change also fixes some draw buffer regressions that happened during
the surface rewrite from the last couple of days.
When using pbuffer or back buffer offscreen rendering the content of
the drawable will have been overwritten. Thus the texture has to be
written into the drawable. An exception is the (pretty common) case
that the whole render target is cleared before rendering to it.
Previously the surfaces stored a flag if the system memory copy was
ahead of the gl copy(SFLAG_DIRTY) or the gl copy is
ahead(SFLAG_GLDIRTY). The pbuffer copy was 'managed' differently using
SFLAG_INPBUFFER and SFLAG_INTEXTURE.
This patch replaces them with 3 flags, INSYSMEM, INPBUFFER and
INTEXTURE which specify which copy contains the most up to date
copy. It is perfectly valid to have more than one of those flags
set. One must be set at least (except at init, when no content is in
the surface yet). When one copy is modified, the flags for the others
are removed.
The method is removed because it does not really help with
anything. It should not be exported from wined3d, there is no need for
the other libs to call it. It does not help abstraction and code
simplification in any way because it is very specific and the code
calling it has to know what is happening in the surface to use this
method.
OpenGL AUX buffers provide a way for offscreen rendering which is very
similar to our back buffer "offscreen" rendering emulation. Not all
card support aux buffers, but if they are available they are a nice
present which is easy to use.
Stefan Dösinger