Should work correctly for all pixel formats except SRGB formats.
Generated code made much simpler by defining the pixel format as
a C structure. For example this is the generated structure for
PIPE_FORMAT_B6UG5SR5S_NORM:
union util_format_b6ug5sr5s_norm {
uint16_t value;
struct {
int r:5;
int g:5;
unsigned b:6;
} chan;
};
Not used everywhere yet because it seems compiled code is slower than
bitshift arithmetic by some misterious reason. So we should generate
bitshift arithmetic at least for the simple UNORM pixel formats.
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.
Currently counting number of tris, how many tiles of each size are
fully covered, partially covered or empty, etc.
Set LP_DEBUG=counters to enable. Results are printed upon context
destruction.
Was used only as a reference, since texture sampling is now code generated.
Already axed in the lp-binning branch too.
This fixes the llvmpipe build after recent sampling changes.
The setup tiling engine is now plugged directly into the draw module
as a rendering backend.
Removed a couple of layering violations such that the setup code no
longer reaches out into the surrounding llvmpipe state or context.
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.
WIP, does't build or run.
Rasterizer code is based on Nick Capen's devmaster posts and the
larrabee articles, but currently doesn't share either the performance
or correctness of either...
This is just a temporary change until we code generate the tile read/write
functions in runtime. The new code avoids an extra memcpy that exists in
u_tile.c functions, from which lp_tile_soa.c was originally based.
This achieves up to 5% improvement, particularly in frames with
little geometry overlap.
Finally a substantial performance improvement: framerates of apps using
texturing tripled, and furthermore, enabling/disabling texturing only
affects around 15% of the framerate, which means the bottleneck is now
somewhere else.
Generated texture sampling code is not complete though -- we always
sample from the base level -- so final figures will be different.
lp_winsys will eventually be unified with softpipe's eventually, but we
are free to move quicker since we don't have the myriad of users yet.
Will provide a pipe_winsys adaptor from Keith's softpipe-private-winsys
soon.
Special attention is given to the interpolation of side by side quads.
Multiplications are made only for the first quad. Interpolation of
inputs for posterior quads are done exclusively with additions, and
perspective divide if necessary.
José Fonseca