We were doing this for the sake of softpipe and the tgsi intergrepter since
we always need the fragment position and W-coordinate information in order
to compute fragment interpolants.
But that's not appropriate for hardware drivers.
The tgsi interpreter now get x,y,w information from a separate tgsi_exec_vector
variable setup by softpipe.
The new pipe_shader_state->input_map[] defines how vert shader outputs map
to frag shader inputs. It may go away though, since one can also examine
the semantic label on frag shader input[0] to figure things out.
Build a buffer of contigous vertices and indices at the backend of our
software transformation/clipping path. This will become the mechanism
for emitting buffers of vertices to rasterization hardware.
This is similar to but not the same as the post-transform vertex cache.
In particular, these vertices are subject to clipping, culling, poly offset,
etc. The vertices emitted will all be used by hardware.
TODOs include the actual transformation to hardware vertex formats, moving
this out of softpipe to somewhere more useful and allowing >1 primitive to
share the generated VB.
This is a step toward removing TGSI_ATTRIB_ tokens.
Basically, when translating Mesa programs to TGSI programs, pass in input and
output register re-maps, plus interpolation info.
There's some known breakage (cubemap.c) so more to be done...
One less dependency on the TGSI_ATTRIB_x flags.
This requires setting the vertex_info->interp_mode[] values in the i915 driver and passing them to draw_set_vertex_attributes().
The code is all in place, but mostly disabled for now:
In t_vp_build.c, write the VERT_RESULT_PSIZE register
In sp_state_derived.c, need to emit vertex point size if drawing points.
In setup_point() use the point size from the vertex.
Replace VF_ATTRIB_x with TGSI_ATTRIB_x
When converting mesa programs to TGSI programs, also convert the InputsRead
and OutputsWritten to a mask of TGSI_ATTRIB_ bits.
Still need to do conversion for vertex programs...
