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