+ assert(lp_check_alignment(color, llvmpipe_get_format_alignment(task->scene->fb.cbufs[buf]->format)));
+ return color;
+}
+
+
+/**
+ * Get the pointer to a 4x4 depth block (within a 64x64 tile).
+ * \param x, y location of 4x4 block in window coords
+ */
+static inline uint8_t *
+lp_rast_get_depth_block_pointer(struct lp_rasterizer_task *task,
+ unsigned x, unsigned y, unsigned layer)
+{
+ unsigned px, py, pixel_offset;
+ uint8_t *depth;
+
+ assert(x < task->scene->tiles_x * TILE_SIZE);
+ assert(y < task->scene->tiles_y * TILE_SIZE);
+ assert((x % TILE_VECTOR_WIDTH) == 0);
+ assert((y % TILE_VECTOR_HEIGHT) == 0);
+
+ assert(task->depth_tile);
+
+ px = x % TILE_SIZE;
+ py = y % TILE_SIZE;
+
+ pixel_offset = px * task->scene->zsbuf.format_bytes +
+ py * task->scene->zsbuf.stride;
+ depth = task->depth_tile + pixel_offset;
+
+ if (layer) {
+ depth += layer * task->scene->zsbuf.layer_stride;
+ }
+
+ assert(lp_check_alignment(depth, llvmpipe_get_format_alignment(task->scene->fb.zsbuf->format)));
+ return depth;
+}
+
+
+
+/**
+ * Shade all pixels in a 4x4 block. The fragment code omits the
+ * triangle in/out tests.
+ * \param x, y location of 4x4 block in window coords
+ */
+static inline void
+lp_rast_shade_quads_all( struct lp_rasterizer_task *task,
+ const struct lp_rast_shader_inputs *inputs,
+ unsigned x, unsigned y )
+{
+ const struct lp_scene *scene = task->scene;
+ const struct lp_rast_state *state = task->state;
+ struct lp_fragment_shader_variant *variant = state->variant;
+ uint8_t *color[PIPE_MAX_COLOR_BUFS];
+ unsigned stride[PIPE_MAX_COLOR_BUFS];
+ uint8_t *depth = NULL;
+ unsigned depth_stride = 0;
+ unsigned i;
+
+ /* color buffer */
+ for (i = 0; i < scene->fb.nr_cbufs; i++) {
+ if (scene->fb.cbufs[i]) {
+ stride[i] = scene->cbufs[i].stride;
+ color[i] = lp_rast_get_color_block_pointer(task, i, x, y,
+ inputs->layer);
+ }
+ else {
+ stride[i] = 0;
+ color[i] = NULL;
+ }
+ }
+
+ if (scene->zsbuf.map) {
+ depth = lp_rast_get_depth_block_pointer(task, x, y, inputs->layer);
+ depth_stride = scene->zsbuf.stride;
+ }
+
+ /*
+ * The rasterizer may produce fragments outside our
+ * allocated 4x4 blocks hence need to filter them out here.
+ */
+ if ((x % TILE_SIZE) < task->width && (y % TILE_SIZE) < task->height) {
+ /* not very accurate would need a popcount on the mask */
+ /* always count this not worth bothering? */
+ task->ps_invocations += 1 * variant->ps_inv_multiplier;
+
+ /* Propagate non-interpolated raster state. */
+ task->thread_data.raster_state.viewport_index = inputs->viewport_index;
+
+ /* run shader on 4x4 block */
+ BEGIN_JIT_CALL(state, task);
+ variant->jit_function[RAST_WHOLE]( &state->jit_context,
+ x, y,
+ inputs->frontfacing,
+ GET_A0(inputs),
+ GET_DADX(inputs),
+ GET_DADY(inputs),
+ color,
+ depth,
+ 0xffff,
+ &task->thread_data,
+ stride,
+ depth_stride);
+ END_JIT_CALL();
+ }
+}
+
+void lp_rast_triangle_1( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_2( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_3( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_4( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_5( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_6( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_7( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_8( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+
+void lp_rast_triangle_3_4(struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+
+void lp_rast_triangle_3_16( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+
+void lp_rast_triangle_4_16( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+
+
+void lp_rast_triangle_32_1( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_32_2( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_32_3( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_32_4( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_32_5( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_32_6( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_32_7( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_32_8( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+
+void lp_rast_triangle_32_3_4(struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+
+void lp_rast_triangle_32_3_16( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );