This is a necessary change to emit the right instructions when writing
to result.depth.
However, even with this test, Z-write doesn't work properly, and I don't
fully understand why. In addition to this, we'll at least have to disable
early-Z, but even that doesn't seem to be enough.
Fix a bug in the LIT implementation (clamp exponent to 128, not 0.5)
and change the implementation around. In theory, the new implementation
needs as little as 5 instruction slots. Unfortunately, the dependency
analysis in find_and_replace_slot is not strong enough to look at
individual components of a register yet.
There were a number of bugs related to the pairing of vector and scalar
operations where swizzles ended up using the wrong source register,
or an instruction was moved forward and ended up overwriting an aliased
register.
The new algorithm for register allocation is quite conservative and may
run out of registers before necessary. On the plus side, It Just Works.
Pairing is done whenever possible, and in more cases than before, so
in practice this change should be a net win.
Remove superfluous _mesa_resize_framebuffer call which is now harmful because
it causes the third renderbuffer to have width/height 0, so Mesa refuses to
render to it.
In the long term, it would be nice to remove the hack in
intel_alloc_window_storage in favour of a proper Mesa interface for flipping
between more than two colour buffers.
In xmesa_check_and_update_buffer_size() handle xmctx==NULL correctly: still
call _mesa_resize_framebufer(). If we don't we can wind up in a situation
where the framebuffer size is non-zero but an attached renderbuffer size
is still initialized to zero. This inconsistancy can later cause problems.
Check for zero-size renderbuffers in update_color_draw_buffers() and
update_color_read_buffer().
See bug 7205.
Reorder fields according to the order in which the pixel transfer operations
take place. Improve comments.
Move the pixel maps out of gl_pixel_attrib since they're not supposed to be
pushed/popped by glPush/PopAttrib.
New gl_pixelmap and gl_pixelmaps structs to contain the pixelmaps.
Bring over the optimizations for fog and normalized spot dir
from t_vp_build.c to brw_vs_tnl.c. Likewise, port a fix for point size
calc from brw_vs_tnl.c to t_vp_build.c (use ABS(eyez) instead of -eyez). Leave
the now differing point size calcs alone though, not sure what's better (it's
basically MOV, ABS, MUL, DP3 vs. ABS, MAD, MAD).
ARB_vertex_buffer_object looks like a useful extension even for old chips.
The drivers should not need any code to be able to use this extension since
they just use mesa's vbo code anyway.
Newly enabled for i810, mach64, mga, r128, radeon, savage, sis and unichrome.
