radeonsi: clean up lucky #include dependencies
[mesa.git] / src / gallium / drivers / radeonsi / si_pipe.h
1 /*
2 * Copyright 2010 Jerome Glisse <glisse@freedesktop.org>
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * on the rights to use, copy, modify, merge, publish, distribute, sub
8 * license, and/or sell copies of the Software, and to permit persons to whom
9 * the Software is furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
13 * Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
19 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
20 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
21 * USE OR OTHER DEALINGS IN THE SOFTWARE.
22 *
23 * Authors:
24 * Jerome Glisse
25 */
26 #ifndef SI_PIPE_H
27 #define SI_PIPE_H
28
29 #include "si_shader.h"
30
31 #ifdef PIPE_ARCH_BIG_ENDIAN
32 #define SI_BIG_ENDIAN 1
33 #else
34 #define SI_BIG_ENDIAN 0
35 #endif
36
37 /* The base vertex and primitive restart can be any number, but we must pick
38 * one which will mean "unknown" for the purpose of state tracking and
39 * the number shouldn't be a commonly-used one. */
40 #define SI_BASE_VERTEX_UNKNOWN INT_MIN
41 #define SI_RESTART_INDEX_UNKNOWN INT_MIN
42 #define SI_NUM_SMOOTH_AA_SAMPLES 8
43 #define SI_GS_PER_ES 128
44
45 /* Instruction cache. */
46 #define SI_CONTEXT_INV_ICACHE (R600_CONTEXT_PRIVATE_FLAG << 0)
47 /* SMEM L1, other names: KCACHE, constant cache, DCACHE, data cache */
48 #define SI_CONTEXT_INV_SMEM_L1 (R600_CONTEXT_PRIVATE_FLAG << 1)
49 /* VMEM L1 can optionally be bypassed (GLC=1). Other names: TC L1 */
50 #define SI_CONTEXT_INV_VMEM_L1 (R600_CONTEXT_PRIVATE_FLAG << 2)
51 /* Used by everything except CB/DB, can be bypassed (SLC=1). Other names: TC L2 */
52 #define SI_CONTEXT_INV_GLOBAL_L2 (R600_CONTEXT_PRIVATE_FLAG << 3)
53 /* Framebuffer caches. */
54 #define SI_CONTEXT_FLUSH_AND_INV_CB_META (R600_CONTEXT_PRIVATE_FLAG << 4)
55 #define SI_CONTEXT_FLUSH_AND_INV_DB_META (R600_CONTEXT_PRIVATE_FLAG << 5)
56 #define SI_CONTEXT_FLUSH_AND_INV_DB (R600_CONTEXT_PRIVATE_FLAG << 6)
57 #define SI_CONTEXT_FLUSH_AND_INV_CB (R600_CONTEXT_PRIVATE_FLAG << 7)
58 /* Engine synchronization. */
59 #define SI_CONTEXT_VS_PARTIAL_FLUSH (R600_CONTEXT_PRIVATE_FLAG << 8)
60 #define SI_CONTEXT_PS_PARTIAL_FLUSH (R600_CONTEXT_PRIVATE_FLAG << 9)
61 #define SI_CONTEXT_CS_PARTIAL_FLUSH (R600_CONTEXT_PRIVATE_FLAG << 10)
62 #define SI_CONTEXT_VGT_FLUSH (R600_CONTEXT_PRIVATE_FLAG << 11)
63 #define SI_CONTEXT_VGT_STREAMOUT_SYNC (R600_CONTEXT_PRIVATE_FLAG << 12)
64
65 #define SI_CONTEXT_FLUSH_AND_INV_FRAMEBUFFER (SI_CONTEXT_FLUSH_AND_INV_CB | \
66 SI_CONTEXT_FLUSH_AND_INV_CB_META | \
67 SI_CONTEXT_FLUSH_AND_INV_DB | \
68 SI_CONTEXT_FLUSH_AND_INV_DB_META)
69
70 #define SI_ENCODE_TRACE_POINT(id) (0xcafe0000 | ((id) & 0xffff))
71 #define SI_IS_TRACE_POINT(x) (((x) & 0xcafe0000) == 0xcafe0000)
72 #define SI_GET_TRACE_POINT_ID(x) ((x) & 0xffff)
73
74 #define SI_MAX_BORDER_COLORS 4096
75
76 struct si_compute;
77 struct hash_table;
78 struct u_suballocator;
79
80 struct si_screen {
81 struct r600_common_screen b;
82 unsigned gs_table_depth;
83 unsigned tess_offchip_block_dw_size;
84 bool has_distributed_tess;
85 bool has_draw_indirect_multi;
86
87 /* Whether shaders are monolithic (1-part) or separate (3-part). */
88 bool use_monolithic_shaders;
89 bool record_llvm_ir;
90
91 pipe_mutex shader_parts_mutex;
92 struct si_shader_part *vs_prologs;
93 struct si_shader_part *vs_epilogs;
94 struct si_shader_part *tcs_epilogs;
95 struct si_shader_part *ps_prologs;
96 struct si_shader_part *ps_epilogs;
97
98 /* Shader cache in memory.
99 *
100 * Design & limitations:
101 * - The shader cache is per screen (= per process), never saved to
102 * disk, and skips redundant shader compilations from TGSI to bytecode.
103 * - It can only be used with one-variant-per-shader support, in which
104 * case only the main (typically middle) part of shaders is cached.
105 * - Only VS, TCS, TES, PS are cached, out of which only the hw VS
106 * variants of VS and TES are cached, so LS and ES aren't.
107 * - GS and CS aren't cached, but it's certainly possible to cache
108 * those as well.
109 */
110 pipe_mutex shader_cache_mutex;
111 struct hash_table *shader_cache;
112
113 /* Shader compiler queue for multithreaded compilation. */
114 struct util_queue shader_compiler_queue;
115 LLVMTargetMachineRef tm[4]; /* used by the queue only */
116 };
117
118 struct si_blend_color {
119 struct r600_atom atom;
120 struct pipe_blend_color state;
121 };
122
123 struct si_sampler_view {
124 struct pipe_sampler_view base;
125 /* [0..7] = image descriptor
126 * [4..7] = buffer descriptor */
127 uint32_t state[8];
128 uint32_t fmask_state[8];
129 const struct radeon_surf_level *base_level_info;
130 unsigned base_level;
131 unsigned block_width;
132 bool is_stencil_sampler;
133 };
134
135 struct si_sampler_state {
136 uint32_t val[4];
137 };
138
139 struct si_cs_shader_state {
140 struct si_compute *program;
141 struct si_compute *emitted_program;
142 unsigned offset;
143 bool initialized;
144 bool uses_scratch;
145 };
146
147 struct si_textures_info {
148 struct si_sampler_views views;
149 uint32_t depth_texture_mask; /* which textures are depth */
150 uint32_t compressed_colortex_mask;
151 };
152
153 struct si_images_info {
154 struct pipe_image_view views[SI_NUM_IMAGES];
155 uint32_t compressed_colortex_mask;
156 unsigned enabled_mask;
157 };
158
159 struct si_framebuffer {
160 struct r600_atom atom;
161 struct pipe_framebuffer_state state;
162 unsigned nr_samples;
163 unsigned log_samples;
164 unsigned cb0_is_integer;
165 unsigned compressed_cb_mask;
166 unsigned spi_shader_col_format;
167 unsigned spi_shader_col_format_alpha;
168 unsigned spi_shader_col_format_blend;
169 unsigned spi_shader_col_format_blend_alpha;
170 unsigned color_is_int8; /* bitmask */
171 unsigned dirty_cbufs;
172 bool dirty_zsbuf;
173 bool any_dst_linear;
174 };
175
176 struct si_clip_state {
177 struct r600_atom atom;
178 struct pipe_clip_state state;
179 };
180
181 struct si_sample_locs {
182 struct r600_atom atom;
183 unsigned nr_samples;
184 };
185
186 struct si_sample_mask {
187 struct r600_atom atom;
188 uint16_t sample_mask;
189 };
190
191 /* A shader state consists of the shader selector, which is a constant state
192 * object shared by multiple contexts and shouldn't be modified, and
193 * the current shader variant selected for this context.
194 */
195 struct si_shader_ctx_state {
196 struct si_shader_selector *cso;
197 struct si_shader *current;
198 };
199
200 struct si_context {
201 struct r600_common_context b;
202 struct blitter_context *blitter;
203 void *custom_dsa_flush;
204 void *custom_blend_resolve;
205 void *custom_blend_decompress;
206 void *custom_blend_fastclear;
207 void *custom_blend_dcc_decompress;
208 struct si_screen *screen;
209
210 struct radeon_winsys_cs *ce_ib;
211 struct radeon_winsys_cs *ce_preamble_ib;
212 bool ce_need_synchronization;
213 struct u_suballocator *ce_suballocator;
214
215 struct si_shader_ctx_state fixed_func_tcs_shader;
216 LLVMTargetMachineRef tm; /* only non-threaded compilation */
217 bool gfx_flush_in_progress;
218 bool compute_is_busy;
219
220 /* Atoms (direct states). */
221 union si_state_atoms atoms;
222 unsigned dirty_atoms; /* mask */
223 /* PM4 states (precomputed immutable states) */
224 union si_state queued;
225 union si_state emitted;
226
227 /* Atom declarations. */
228 struct si_framebuffer framebuffer;
229 struct si_sample_locs msaa_sample_locs;
230 struct r600_atom db_render_state;
231 struct r600_atom msaa_config;
232 struct si_sample_mask sample_mask;
233 struct r600_atom cb_render_state;
234 struct si_blend_color blend_color;
235 struct r600_atom clip_regs;
236 struct si_clip_state clip_state;
237 struct si_shader_data shader_userdata;
238 struct si_stencil_ref stencil_ref;
239 struct r600_atom spi_map;
240
241 /* Precomputed states. */
242 struct si_pm4_state *init_config;
243 struct si_pm4_state *init_config_gs_rings;
244 bool init_config_has_vgt_flush;
245 struct si_pm4_state *vgt_shader_config[4];
246
247 /* shaders */
248 struct si_shader_ctx_state ps_shader;
249 struct si_shader_ctx_state gs_shader;
250 struct si_shader_ctx_state vs_shader;
251 struct si_shader_ctx_state tcs_shader;
252 struct si_shader_ctx_state tes_shader;
253 struct si_cs_shader_state cs_shader_state;
254
255 /* shader information */
256 struct si_vertex_element *vertex_elements;
257 unsigned sprite_coord_enable;
258 bool flatshade;
259 bool do_update_shaders;
260
261 /* shader descriptors */
262 struct si_descriptors vertex_buffers;
263 struct si_descriptors descriptors[SI_NUM_DESCS];
264 unsigned descriptors_dirty;
265 struct si_buffer_resources rw_buffers;
266 struct si_buffer_resources const_buffers[SI_NUM_SHADERS];
267 struct si_buffer_resources shader_buffers[SI_NUM_SHADERS];
268 struct si_textures_info samplers[SI_NUM_SHADERS];
269 struct si_images_info images[SI_NUM_SHADERS];
270
271 /* other shader resources */
272 struct pipe_constant_buffer null_const_buf; /* used for set_constant_buffer(NULL) on CIK */
273 struct pipe_resource *esgs_ring;
274 struct pipe_resource *gsvs_ring;
275 struct pipe_resource *tf_ring;
276 struct pipe_resource *tess_offchip_ring;
277 union pipe_color_union *border_color_table; /* in CPU memory, any endian */
278 struct r600_resource *border_color_buffer;
279 union pipe_color_union *border_color_map; /* in VRAM (slow access), little endian */
280 unsigned border_color_count;
281
282 /* Vertex and index buffers. */
283 bool vertex_buffers_dirty;
284 struct pipe_index_buffer index_buffer;
285 struct pipe_vertex_buffer vertex_buffer[SI_NUM_VERTEX_BUFFERS];
286
287 /* MSAA config state. */
288 int ps_iter_samples;
289 bool smoothing_enabled;
290
291 /* DB render state. */
292 bool dbcb_depth_copy_enabled;
293 bool dbcb_stencil_copy_enabled;
294 unsigned dbcb_copy_sample;
295 bool db_flush_depth_inplace;
296 bool db_flush_stencil_inplace;
297 bool db_depth_clear;
298 bool db_depth_disable_expclear;
299 bool db_stencil_clear;
300 bool db_stencil_disable_expclear;
301 unsigned ps_db_shader_control;
302 bool occlusion_queries_disabled;
303
304 /* Emitted draw state. */
305 int last_index_size;
306 int last_base_vertex;
307 int last_start_instance;
308 int last_drawid;
309 int last_sh_base_reg;
310 int last_primitive_restart_en;
311 int last_restart_index;
312 int last_gs_out_prim;
313 int last_prim;
314 int last_multi_vgt_param;
315 int last_rast_prim;
316 unsigned last_sc_line_stipple;
317 int last_vtx_reuse_depth;
318 int current_rast_prim; /* primitive type after TES, GS */
319 unsigned last_gsvs_itemsize;
320
321 /* Scratch buffer */
322 struct r600_resource *scratch_buffer;
323 bool emit_scratch_reloc;
324 unsigned scratch_waves;
325 unsigned spi_tmpring_size;
326
327 struct r600_resource *compute_scratch_buffer;
328
329 /* Emitted derived tessellation state. */
330 struct si_shader *last_ls; /* local shader (VS) */
331 struct si_shader_selector *last_tcs;
332 int last_num_tcs_input_cp;
333 int last_tes_sh_base;
334
335 /* Debug state. */
336 bool is_debug;
337 struct radeon_saved_cs last_gfx;
338 struct r600_resource *last_trace_buf;
339 struct r600_resource *trace_buf;
340 unsigned trace_id;
341 uint64_t dmesg_timestamp;
342 unsigned apitrace_call_number;
343
344 /* Other state */
345 bool need_check_render_feedback;
346 };
347
348 /* cik_sdma.c */
349 void cik_init_sdma_functions(struct si_context *sctx);
350
351 /* si_blit.c */
352 void si_init_blit_functions(struct si_context *sctx);
353 void si_decompress_graphics_textures(struct si_context *sctx);
354 void si_decompress_compute_textures(struct si_context *sctx);
355 void si_resource_copy_region(struct pipe_context *ctx,
356 struct pipe_resource *dst,
357 unsigned dst_level,
358 unsigned dstx, unsigned dsty, unsigned dstz,
359 struct pipe_resource *src,
360 unsigned src_level,
361 const struct pipe_box *src_box);
362
363 /* si_cp_dma.c */
364 void si_copy_buffer(struct si_context *sctx,
365 struct pipe_resource *dst, struct pipe_resource *src,
366 uint64_t dst_offset, uint64_t src_offset, unsigned size);
367 void si_init_cp_dma_functions(struct si_context *sctx);
368
369 /* si_debug.c */
370 void si_init_debug_functions(struct si_context *sctx);
371 void si_check_vm_faults(struct r600_common_context *ctx,
372 struct radeon_saved_cs *saved, enum ring_type ring);
373 bool si_replace_shader(unsigned num, struct radeon_shader_binary *binary);
374
375 /* si_dma.c */
376 void si_init_dma_functions(struct si_context *sctx);
377
378 /* si_hw_context.c */
379 void si_context_gfx_flush(void *context, unsigned flags,
380 struct pipe_fence_handle **fence);
381 void si_begin_new_cs(struct si_context *ctx);
382 void si_need_cs_space(struct si_context *ctx);
383
384 /* si_compute.c */
385 void si_init_compute_functions(struct si_context *sctx);
386
387 /* si_perfcounters.c */
388 void si_init_perfcounters(struct si_screen *screen);
389
390 /* si_uvd.c */
391 struct pipe_video_codec *si_uvd_create_decoder(struct pipe_context *context,
392 const struct pipe_video_codec *templ);
393
394 struct pipe_video_buffer *si_video_buffer_create(struct pipe_context *pipe,
395 const struct pipe_video_buffer *tmpl);
396
397 /*
398 * common helpers
399 */
400
401 static inline struct r600_resource *
402 si_resource_create_custom(struct pipe_screen *screen,
403 unsigned usage, unsigned size)
404 {
405 assert(size);
406 return r600_resource(pipe_buffer_create(screen,
407 PIPE_BIND_CUSTOM, usage, size));
408 }
409
410 static inline void
411 si_invalidate_draw_sh_constants(struct si_context *sctx)
412 {
413 sctx->last_base_vertex = SI_BASE_VERTEX_UNKNOWN;
414 }
415
416 static inline void
417 si_set_atom_dirty(struct si_context *sctx,
418 struct r600_atom *atom, bool dirty)
419 {
420 unsigned bit = 1 << (atom->id - 1);
421
422 if (dirty)
423 sctx->dirty_atoms |= bit;
424 else
425 sctx->dirty_atoms &= ~bit;
426 }
427
428 static inline bool
429 si_is_atom_dirty(struct si_context *sctx,
430 struct r600_atom *atom)
431 {
432 unsigned bit = 1 << (atom->id - 1);
433
434 return sctx->dirty_atoms & bit;
435 }
436
437 static inline void
438 si_mark_atom_dirty(struct si_context *sctx,
439 struct r600_atom *atom)
440 {
441 si_set_atom_dirty(sctx, atom, true);
442 }
443
444 static inline struct tgsi_shader_info *si_get_vs_info(struct si_context *sctx)
445 {
446 if (sctx->gs_shader.cso)
447 return &sctx->gs_shader.cso->info;
448 else if (sctx->tes_shader.cso)
449 return &sctx->tes_shader.cso->info;
450 else if (sctx->vs_shader.cso)
451 return &sctx->vs_shader.cso->info;
452 else
453 return NULL;
454 }
455
456 static inline struct si_shader* si_get_vs_state(struct si_context *sctx)
457 {
458 if (sctx->gs_shader.current)
459 return sctx->gs_shader.current->gs_copy_shader;
460 else if (sctx->tes_shader.current)
461 return sctx->tes_shader.current;
462 else
463 return sctx->vs_shader.current;
464 }
465
466 static inline bool si_vs_exports_prim_id(struct si_shader *shader)
467 {
468 if (shader->selector->type == PIPE_SHADER_VERTEX)
469 return shader->key.vs.epilog.export_prim_id;
470 else if (shader->selector->type == PIPE_SHADER_TESS_EVAL)
471 return shader->key.tes.epilog.export_prim_id;
472 else
473 return false;
474 }
475
476 #endif