1 /**************************************************************************
3 * Copyright 2009 VMware, Inc.
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 **************************************************************************/
28 #ifndef LP_RAST_PRIV_H
29 #define LP_RAST_PRIV_H
31 #include "os/os_thread.h"
32 #include "util/u_format.h"
33 #include "gallivm/lp_bld_debug.h"
34 #include "lp_memory.h"
38 #include "lp_texture.h"
39 #include "lp_tile_soa.h"
40 #include "lp_limits.h"
43 /* If we crash in a jitted function, we can examine jit_line and jit_state
44 * to get some info. This is not thread-safe, however.
48 struct lp_rasterizer_task;
50 extern const struct lp_rast_state *jit_state;
51 extern const struct lp_rasterizer_task *jit_task;
53 #define BEGIN_JIT_CALL(state, task) \
55 jit_line = __LINE__; \
60 #define END_JIT_CALL() \
68 #define BEGIN_JIT_CALL(X, Y)
69 #define END_JIT_CALL()
78 * Per-thread rasterization state
80 struct lp_rasterizer_task
82 const struct cmd_bin *bin;
83 const struct lp_rast_state *state;
85 struct lp_scene *scene;
86 unsigned x, y; /**< Pos of this tile in framebuffer, in pixels */
88 uint8_t *color_tiles[PIPE_MAX_COLOR_BUFS];
92 struct lp_rasterizer *rast;
95 unsigned thread_index;
97 /* occlude counter for visiable pixels */
99 struct llvmpipe_query *query;
101 pipe_semaphore work_ready;
102 pipe_semaphore work_done;
107 * This is the state required while rasterizing tiles.
108 * Note that this contains per-thread information too.
109 * The tile size is TILE_SIZE x TILE_SIZE pixels.
115 /** The incoming queue of scenes ready to rasterize */
116 struct lp_scene_queue *full_scenes;
118 /** The scene currently being rasterized by the threads */
119 struct lp_scene *curr_scene;
121 /** A task object for each rasterization thread */
122 struct lp_rasterizer_task tasks[LP_MAX_THREADS];
124 unsigned num_threads;
125 pipe_thread threads[LP_MAX_THREADS];
127 /** For synchronizing the rasterization threads */
128 pipe_barrier barrier;
133 lp_rast_shade_quads_mask(struct lp_rasterizer_task *task,
134 const struct lp_rast_shader_inputs *inputs,
135 unsigned x, unsigned y,
141 * Get the pointer to a 4x4 depth/stencil block.
142 * We'll map the z/stencil buffer on demand here.
143 * Note that this may be called even when there's no z/stencil buffer - return
145 * \param x, y location of 4x4 block in window coords
148 lp_rast_get_depth_block_pointer(struct lp_rasterizer_task *task,
149 unsigned x, unsigned y)
151 const struct lp_scene *scene = task->scene;
154 assert((x % TILE_VECTOR_WIDTH) == 0);
155 assert((y % TILE_VECTOR_HEIGHT) == 0);
157 if (!scene->zsbuf.map) {
158 /* Either out of memory or no zsbuf. Can't tell without access
159 * to the state. Just use dummy tile memory, but don't print
160 * the oom warning as this most likely because there is no
163 return lp_dummy_tile;
166 depth = (scene->zsbuf.map +
167 scene->zsbuf.stride * y +
168 scene->zsbuf.blocksize * x * TILE_VECTOR_HEIGHT);
170 assert(lp_check_alignment(depth, 16));
176 * Get pointer to the swizzled color tile
178 static INLINE uint8_t *
179 lp_rast_get_color_tile_pointer(struct lp_rasterizer_task *task,
180 unsigned buf, enum lp_texture_usage usage)
182 const struct lp_scene *scene = task->scene;
184 assert(task->x % TILE_SIZE == 0);
185 assert(task->y % TILE_SIZE == 0);
186 assert(buf < scene->fb.nr_cbufs);
188 if (!task->color_tiles[buf]) {
189 struct pipe_surface *cbuf = scene->fb.cbufs[buf];
190 struct llvmpipe_resource *lpt;
192 lpt = llvmpipe_resource(cbuf->texture);
193 task->color_tiles[buf] = lp_swizzled_cbuf[task->thread_index][buf];
195 if (usage != LP_TEX_USAGE_WRITE_ALL) {
196 llvmpipe_swizzle_cbuf_tile(lpt,
197 cbuf->u.tex.first_layer,
200 task->color_tiles[buf]);
204 return task->color_tiles[buf];
209 * Get the pointer to a 4x4 color block (within a 64x64 tile).
210 * We'll map the color buffer on demand here.
211 * Note that this may be called even when there's no color buffers - return
213 * \param x, y location of 4x4 block in window coords
215 static INLINE uint8_t *
216 lp_rast_get_color_block_pointer(struct lp_rasterizer_task *task,
217 unsigned buf, unsigned x, unsigned y)
219 unsigned px, py, pixel_offset;
222 assert((x % TILE_VECTOR_WIDTH) == 0);
223 assert((y % TILE_VECTOR_HEIGHT) == 0);
225 color = lp_rast_get_color_tile_pointer(task, buf, LP_TEX_USAGE_READ_WRITE);
230 pixel_offset = tile_pixel_offset(px, py, 0);
232 color = color + pixel_offset;
234 assert(lp_check_alignment(color, 16));
241 * Shade all pixels in a 4x4 block. The fragment code omits the
242 * triangle in/out tests.
243 * \param x, y location of 4x4 block in window coords
246 lp_rast_shade_quads_all( struct lp_rasterizer_task *task,
247 const struct lp_rast_shader_inputs *inputs,
248 unsigned x, unsigned y )
250 const struct lp_scene *scene = task->scene;
251 const struct lp_rast_state *state = task->state;
252 struct lp_fragment_shader_variant *variant = state->variant;
253 uint8_t *color[PIPE_MAX_COLOR_BUFS];
258 for (i = 0; i < scene->fb.nr_cbufs; i++)
259 color[i] = lp_rast_get_color_block_pointer(task, i, x, y);
261 depth = lp_rast_get_depth_block_pointer(task, x, y);
263 /* run shader on 4x4 block */
264 BEGIN_JIT_CALL(state, task);
265 variant->jit_function[RAST_WHOLE]( &state->jit_context,
274 &task->vis_counter );
278 void lp_rast_triangle_1( struct lp_rasterizer_task *,
279 const union lp_rast_cmd_arg );
280 void lp_rast_triangle_2( struct lp_rasterizer_task *,
281 const union lp_rast_cmd_arg );
282 void lp_rast_triangle_3( struct lp_rasterizer_task *,
283 const union lp_rast_cmd_arg );
284 void lp_rast_triangle_4( struct lp_rasterizer_task *,
285 const union lp_rast_cmd_arg );
286 void lp_rast_triangle_5( struct lp_rasterizer_task *,
287 const union lp_rast_cmd_arg );
288 void lp_rast_triangle_6( struct lp_rasterizer_task *,
289 const union lp_rast_cmd_arg );
290 void lp_rast_triangle_7( struct lp_rasterizer_task *,
291 const union lp_rast_cmd_arg );
292 void lp_rast_triangle_8( struct lp_rasterizer_task *,
293 const union lp_rast_cmd_arg );
295 void lp_rast_triangle_3_4(struct lp_rasterizer_task *,
296 const union lp_rast_cmd_arg );
298 void lp_rast_triangle_3_16( struct lp_rasterizer_task *,
299 const union lp_rast_cmd_arg );
301 void lp_rast_triangle_4_16( struct lp_rasterizer_task *,
302 const union lp_rast_cmd_arg );
305 lp_rast_set_state(struct lp_rasterizer_task *task,
306 const union lp_rast_cmd_arg arg);
309 lp_debug_bin( const struct cmd_bin *bin );