nvfx: rewrite draw code and buffer code

This is a full rewrite of the drawing and buffer management logic.

It offers a lot of improvements:
1. A copy of buffers is now always kept in system memory. This is
   necessary to allow software processing of them, which is necessary
   or improves performance in many cases.
2. Support for pushing vertices on the FIFO, with index lookup if necessary.
3. "Smart" draw code that tries to intelligently choose the cheapest
  way to draw something: whether to use inline vertices or hardware
  vertex buffer, and whether to use hardware index buffers
4. Support for all vertex formats supported by the hardware
5. Usage of translate to push vertices, supporting all formats that are
   sensible to use as vertex formats
6. Support for base vertex
7. Usage of Ben Skeggs' primitive splitter originally for nv50, allowing
   correct splitting of line loops, triangle fans, etc.
8. Support for instancing
9. Precomputation using the vertex elements CSO

Thanks to Ben Skeggs for his primitive splitter originally for nv50.

Thanks to Christoph Bumiller for his nv50 push code, that was the basis
of this work, even though I changed his code dramatically, in particular
to replace his ad-hoc vertex data emitter with translate.

The changes could also go into nv50 too, but there are substantial
differences due to the additional nv50 hardware features.

1 files changed, 90 insertions, 9 deletions

diff --git a/src/gallium/drivers/nvfx/nvfx_context.h b/src/gallium/drivers/nvfx/nvfx_context.h
index bce19df044d..8899bf991e1 100644
--- a/src/gallium/drivers/nvfx/nvfx_context.h
+++ b/src/gallium/drivers/nvfx/nvfx_context.h
@@ -44,6 +44,7 @@
 #define NVFX_NEW_SR		(1 << 13)
 #define NVFX_NEW_VERTCONST	(1 << 14)
 #define NVFX_NEW_FRAGCONST	(1 << 15)
+#define NVFX_NEW_INDEX	(1 << 16)
 
 struct nvfx_rasterizer_state {
 	struct pipe_rasterizer_state pipe;
@@ -71,9 +72,53 @@ struct nvfx_state {
 	unsigned render_temps;
 };
 
+struct nvfx_per_vertex_element {
+	unsigned idx;
+        unsigned vertex_buffer_index;
+        unsigned src_offset;
+};
+
+struct nvfx_low_frequency_element {
+	unsigned idx;
+	unsigned vertex_buffer_index;
+	unsigned src_offset;
+        void (*fetch_rgba_float)(float *dst, const uint8_t *src, unsigned i, unsigned j);
+        unsigned ncomp;
+};
+
+struct nvfx_per_instance_element {
+	struct nvfx_low_frequency_element base;
+	unsigned instance_divisor;
+};
+
+struct nvfx_per_vertex_buffer_info
+{
+	unsigned vertex_buffer_index;
+	unsigned per_vertex_size;
+};
+
 struct nvfx_vtxelt_state {
 	struct pipe_vertex_element pipe[16];
 	unsigned num_elements;
+	unsigned vtxfmt[16];
+
+	unsigned num_per_vertex_buffer_infos;
+	struct nvfx_per_vertex_buffer_info per_vertex_buffer_info[16];
+
+	unsigned num_per_vertex;
+	struct nvfx_per_vertex_element per_vertex[16];
+
+	unsigned num_per_instance;
+	struct nvfx_per_instance_element per_instance[16];
+
+	unsigned num_constant;
+	struct nvfx_low_frequency_element constant[16];
+
+	boolean needs_translate;
+	struct translate* translate;
+
+	unsigned vertex_length;
+	unsigned max_vertices_per_packet;
 };
 
 struct nvfx_render_target {
@@ -127,8 +172,6 @@ struct nvfx_context {
 	struct pipe_viewport_state viewport;
 	struct pipe_framebuffer_state framebuffer;
 	struct pipe_index_buffer idxbuf;
-	struct pipe_resource *idxbuf_buffer;
-	unsigned idxbuf_format;
 	struct nvfx_sampler_state *tex_sampler[PIPE_MAX_SAMPLERS];
 	struct pipe_sampler_view *fragment_sampler_views[PIPE_MAX_SAMPLERS];
 	unsigned nr_samplers;
@@ -137,8 +180,14 @@ struct nvfx_context {
 	struct pipe_vertex_buffer vtxbuf[PIPE_MAX_ATTRIBS];
 	unsigned vtxbuf_nr;
 	struct nvfx_vtxelt_state *vtxelt;
+	int base_vertex;
+	boolean use_index_buffer;
+	/* -1 = hardware input setup is outdated
+	 * 0 = hardware input setup is for inline vertices
+	 * 1 = hardware input setup is for hardware vertices
+	 */
+	int use_vertex_buffers;
 
-	unsigned vbo_bo;
 	unsigned hw_vtxelt_nr;
 	uint8_t hw_samplers;
 	uint32_t hw_txf[8];
@@ -180,11 +229,7 @@ extern void nvfx_clear(struct pipe_context *pipe, unsigned buffers,
 
 /* nvfx_draw.c */
 extern struct draw_stage *nvfx_draw_render_stage(struct nvfx_context *nvfx);
-extern void nvfx_draw_elements_swtnl(struct pipe_context *pipe,
-                                     struct pipe_resource *idxbuf,
-                                     unsigned ib_size, int ib_bias,
-                                     unsigned mode,
-                                     unsigned start, unsigned count);
+extern void nvfx_draw_vbo_swtnl(struct pipe_context *pipe, const struct pipe_draw_info* info);
 extern void nvfx_vtxfmt_validate(struct nvfx_context *nvfx);
 
 /* nvfx_fb.c */
@@ -245,17 +290,53 @@ extern boolean nvfx_state_validate_swtnl(struct nvfx_context *nvfx);
 extern void nvfx_state_emit(struct nvfx_context *nvfx);
 
 /* nvfx_transfer.c */
-extern void nvfx_init_transfer_functions(struct nvfx_context *nvfx);
+extern void nvfx_init_transfer_functions(struct pipe_context *pipe);
 
 /* nvfx_vbo.c */
 extern boolean nvfx_vbo_validate(struct nvfx_context *nvfx);
 extern void nvfx_vbo_relocate(struct nvfx_context *nvfx);
+extern void nvfx_idxbuf_validate(struct nvfx_context* nvfx);
+extern void nvfx_idxbuf_relocate(struct nvfx_context* nvfx);
 extern void nvfx_draw_vbo(struct pipe_context *pipe,
                           const struct pipe_draw_info *info);
+extern void nvfx_init_vbo_functions(struct nvfx_context *nvfx);
+extern unsigned nvfx_vertex_formats[];
 
 /* nvfx_vertprog.c */
 extern boolean nvfx_vertprog_validate(struct nvfx_context *nvfx);
 extern void nvfx_vertprog_destroy(struct nvfx_context *,
 				  struct nvfx_vertex_program *);
 
+/* nvfx_push.c */
+extern void nvfx_push_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info);
+
+/* must WAIT_RING(chan, ncomp + 1) or equivalent beforehand! */
+static inline void nvfx_emit_vtx_attr(struct nouveau_channel* chan, unsigned attrib, float* v, unsigned ncomp)
+{
+	switch (ncomp) {
+	case 4:
+		OUT_RING(chan, RING_3D(NV34TCL_VTX_ATTR_4F_X(attrib), 4));
+		OUT_RING(chan, fui(v[0]));
+		OUT_RING(chan, fui(v[1]));
+		OUT_RING(chan,  fui(v[2]));
+		OUT_RING(chan,  fui(v[3]));
+		break;
+	case 3:
+		OUT_RING(chan, RING_3D(NV34TCL_VTX_ATTR_3F_X(attrib), 3));
+		OUT_RING(chan,  fui(v[0]));
+		OUT_RING(chan,  fui(v[1]));
+		OUT_RING(chan,  fui(v[2]));
+		break;
+	case 2:
+		OUT_RING(chan, RING_3D(NV34TCL_VTX_ATTR_2F_X(attrib), 2));
+		OUT_RING(chan,  fui(v[0]));
+		OUT_RING(chan,  fui(v[1]));
+		break;
+	case 1:
+		OUT_RING(chan, RING_3D(NV34TCL_VTX_ATTR_1F(attrib), 1));
+		OUT_RING(chan,  fui(v[0]));
+		break;
+	}
+}
+
 #endif