Age | Commit message (Collapse) | Author |
|
This fixes several assertion failures due to only Z32 being supported.
|
|
|
|
the llvmpipe tgsi translation is a lot more complete than what was in
gallivm so replacing the latter with the former. this is needed since
the draw llvm paths will use the same code. effectively the proven
llvmpipe code becomes gallivm.
|
|
|
|
Conflicts:
src/gallium/auxiliary/util/u_dl.c
src/gallium/auxiliary/util/u_time.h
src/gallium/drivers/llvmpipe/lp_state_derived.c
src/gallium/drivers/llvmpipe/lp_state_surface.c
src/gallium/drivers/llvmpipe/lp_tex_cache.c
src/gallium/drivers/llvmpipe/lp_tile_cache.c
|
|
The surface is always a multiple of the tile size now.
|
|
|
|
|
|
These options draw lines over the tiles to show the 64x64 tile
bounds and 16x16 sub-tile bounds. For debugging/visualization.
|
|
|
|
When we know that a 4x4 pixel block is entirely inside of a triangle
use the jit function which omits the in/out test code.
Results in a few percent speedup in many tests.
|
|
This is just a tiny bit faster.
|
|
This makes a big difference in progs that clear to a non-gray color.
Some demos are 30-50% faster.
|
|
Adjust definition of empty_bin according to what's actually in empty
bins. We often have a state packet before/after load commands.
Still need to do something about the fence packets.
|
|
Load zbuffer contents for binned scenes that don't start with a clear
and which have a bound zbuffer.
|
|
Remove unused param, add comments. Thanks to Brian for review.
|
|
We emit at most two clear packets (color and z respectively).
|
|
|
|
Non-mrt apps work, and the code looks correct, but not many mrt test apps
handy atm...
|
|
Conflicts:
src/gallium/auxiliary/util/u_surface.c
src/gallium/drivers/llvmpipe/Makefile
src/gallium/drivers/llvmpipe/SConscript
src/gallium/drivers/llvmpipe/lp_bld_arit.c
src/gallium/drivers/llvmpipe/lp_bld_flow.c
src/gallium/drivers/llvmpipe/lp_bld_interp.c
src/gallium/drivers/llvmpipe/lp_clear.c
src/gallium/drivers/llvmpipe/lp_context.c
src/gallium/drivers/llvmpipe/lp_context.h
src/gallium/drivers/llvmpipe/lp_draw_arrays.c
src/gallium/drivers/llvmpipe/lp_jit.c
src/gallium/drivers/llvmpipe/lp_jit.h
src/gallium/drivers/llvmpipe/lp_prim_vbuf.c
src/gallium/drivers/llvmpipe/lp_setup.c
src/gallium/drivers/llvmpipe/lp_setup_point.c
src/gallium/drivers/llvmpipe/lp_state.h
src/gallium/drivers/llvmpipe/lp_state_blend.c
src/gallium/drivers/llvmpipe/lp_state_derived.c
src/gallium/drivers/llvmpipe/lp_state_fs.c
src/gallium/drivers/llvmpipe/lp_state_sampler.c
src/gallium/drivers/llvmpipe/lp_state_surface.c
src/gallium/drivers/llvmpipe/lp_tex_cache.c
src/gallium/drivers/llvmpipe/lp_tex_cache.h
src/gallium/drivers/llvmpipe/lp_tex_sample.h
src/gallium/drivers/llvmpipe/lp_tile_cache.c
|
|
|
|
Since changing the in/out test we can just use INT_MIN to be sure the
comparison against the step values always passes.
|
|
Instead of:
s = c + step
m = s > 0
Do:
m = step > c (with negated c)
|
|
The test to determine which of the pixels in a 2x2 quad is now done in
the fragment shader rather than in the calling C code. This is a little
faster but there's a few more things to do.
Note that the step[] array elements are in a different order now. Rather
than being in row-major order for the 4x4 grid, they're in "quad-major"
order. The setup of the step arrays is a little more complicated now.
So is the course/intermediate tile test code, but some lookup tables
help with that.
Next steps:
- early-cull 2x2 quads which are totally outside the triangle.
- skip the in/out test for fully contained quads
- make the in/out comparison code tighter/faster.
|
|
|
|
It was pretty confusing having an entity named "bin" and another named
"bins", not least because sometimes there was a need to talk about >1
of the "bins" objects, which couldn't be pluralized any further...
Scene is a term used in a bunch of places to talk about what a binner
operates on, so it's a decent choice here.
|
|
|
|
|
|
|
|
Now mapping/unmapping the framebuffer is done by a rasteizer thread
rather than the main calling thread.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Also bump MAX_THREADS to 8.
|
|
|
|
The LP_NUM_THREADS env var controls how many threads are created.
The default (and max) is 4, for now.
If LP_NUM_THREADS = 0, threading is not used.
|
|
Some of the state is per-thread. Put that state in new lp_rasterizer_task
struct.
|
|
|
|
Move tiles_x,y fields from setup state into bin state.
Move more bin-adding commands into lp_bin.[ch].
|
|
New lp_bins struct contains all bin information.
More move bin-related code into lp_bin.[ch]
Use new/updated bin-access functions to hide implementation details.
The result is more/cleaner separation between the setup and rast components.
This will make double-buffering of the bins easier, etc.
|
|
First step of moving bin rasterization/execution code out of lp_setup.c
|
|
|
|
Previously, each triangle had a pointer to the state to use for shading.
Now we insert state-change commands into the bins. When we execute one
of those commands we just update a 'current state' pointer and use that
pointer when calling the jit shader.
When inserting state-change commands into a bin we check if the previous
command was also a state-change command and simply replace it. This
avoids accumulating useless/redundant state-change commands.
|
|
|
|
|