The idea behind this is that anything touching registers should be in
r600_state.c or evergreen_state.c. This is also consistent with
evergreen_pipe_shader_vs().
Signed-off-by: Henri Verbeet <hverbeet@gmail.com>
We need more and more of these, and it is difficult and prone to version
incompatability issues trying to single out every one of them.
This mimicks what was done in SCons.
This enables the new shadow texture functions in GLSL 1.30.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chad Versace <chad.versace@intel.com>
The assignment on line 368, `tex_swizzles[i] = SWIZZLE_NOOP`, is rendered
dead by the reassignment on line 392.
Signed-off-by: Chad Versace <chad.versace@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This is necessary for GLSL 1.30+ shadow sampling functions, which return
a single float rather than splatting the value to a vec4 based on
GL_DEPTH_TEXTURE_MODE.
This was probably missed when implementing luminance and luminance alpha
render targets.
_mesa_get_format_bits checks for both GL_*_BITS and GL_TEXTURE_*_SIZE.
This fixes:
main/framebuffer.c:892: _mesa_source_buffer_exists: Assertion `....' failed.
The r300 compiler can eliminate unused uniforms and remap uniform locations
if their number surpasses hardware limits, so the limit is actually
NumParameters + NumUnusedParameters. This is important for some apps
under Wine to run.
Wine sometimes declares a uniform array of 256 vec4's and some Wine-specific
constants on top of that, so in total there is more uniforms than r300 can
handle. This was the main motivation for implementing the elimination
of unused constants.
We should allow drivers to implement fail & recovery paths where it makes
sense, so giving up too early especially when comes to uniforms is not
so good idea, though I agree there should be some hard limit for all drivers.
This patch fixes:
- glsl-fs-uniform-array-5
- glsl-vs-large-uniform-array
on drivers which can eliminate unused uniforms.
Avoid setting the same gpu register several times in a r600_pipe_state.
Compute the final value of the register and set that one time. This avoids
some overhead in r600_context_pipe_state_set().
Signed-off-by: Mathias Fröhlich <Mathias.Froehlich@web.de>
Signed-off-by: Henri Verbeet <hverbeet@gmail.com>
Well, not sure what exactly it is, but it certainly doesn't contain
the control flow stack, but vertex data.
Not sure about size, I've only seen the first few KiB written, but
the binary driver seems to allocate more.
Depth values are also written before the shader is executed, so if
early tests are enabled, fragments that failed the alpha test were
modifying the depth buffer, but they shouldn't.
The kernel drm takes care of all coherency as long as we don't forget
to submit all outstanding commands in the batchbuffer ...
Also move batchbuffer initialization up because otherwise transfers
for some helper textures fail with a segmentation fault.
And kill the dead code, flushes should now be correct everywhere.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>