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 R600_PIPE_H
27 #define R600_PIPE_H
28 
29 #include "r600_pipe_common.h"
30 #include "r600_cs.h"
31 #include "r600_public.h"
32 #include "pipe/p_defines.h"
33 
34 #include "util/u_suballoc.h"
35 #include "util/list.h"
36 #include "util/u_transfer.h"
37 #include "util/u_memory.h"
38 
39 #include "tgsi/tgsi_scan.h"
40 
41 #define R600_NUM_ATOMS 56
42 
43 #define R600_MAX_IMAGES 8
44 /*
45  * ranges reserved for images on evergreen
46  * first set for the immediate buffers,
47  * second for the actual resources for RESQ.
48  */
49 #define R600_IMAGE_IMMED_RESOURCE_OFFSET 160
50 #define R600_IMAGE_REAL_RESOURCE_OFFSET 168
51 
52 /* read caches */
53 #define R600_CONTEXT_INV_VERTEX_CACHE		(R600_CONTEXT_PRIVATE_FLAG << 0)
54 #define R600_CONTEXT_INV_TEX_CACHE		(R600_CONTEXT_PRIVATE_FLAG << 1)
55 #define R600_CONTEXT_INV_CONST_CACHE		(R600_CONTEXT_PRIVATE_FLAG << 2)
56 /* read-write caches */
57 #define R600_CONTEXT_FLUSH_AND_INV		(R600_CONTEXT_PRIVATE_FLAG << 3)
58 #define R600_CONTEXT_FLUSH_AND_INV_CB_META	(R600_CONTEXT_PRIVATE_FLAG << 4)
59 #define R600_CONTEXT_FLUSH_AND_INV_DB_META	(R600_CONTEXT_PRIVATE_FLAG << 5)
60 #define R600_CONTEXT_FLUSH_AND_INV_DB		(R600_CONTEXT_PRIVATE_FLAG << 6)
61 #define R600_CONTEXT_FLUSH_AND_INV_CB		(R600_CONTEXT_PRIVATE_FLAG << 7)
62 /* engine synchronization */
63 #define R600_CONTEXT_PS_PARTIAL_FLUSH		(R600_CONTEXT_PRIVATE_FLAG << 8)
64 #define R600_CONTEXT_WAIT_3D_IDLE		(R600_CONTEXT_PRIVATE_FLAG << 9)
65 #define R600_CONTEXT_WAIT_CP_DMA_IDLE		(R600_CONTEXT_PRIVATE_FLAG << 10)
66 #define R600_CONTEXT_CS_PARTIAL_FLUSH           (R600_CONTEXT_PRIVATE_FLAG << 11)
67 
68 /* the number of CS dwords for flushing and drawing */
69 #define R600_MAX_FLUSH_CS_DWORDS	18
70 #define R600_MAX_DRAW_CS_DWORDS		58
71 #define R600_MAX_PFP_SYNC_ME_DWORDS	16
72 
73 #define EG_MAX_ATOMIC_BUFFERS 8
74 
75 #define R600_MAX_USER_CONST_BUFFERS 15
76 #define R600_MAX_DRIVER_CONST_BUFFERS 3
77 #define R600_MAX_CONST_BUFFERS (R600_MAX_USER_CONST_BUFFERS + R600_MAX_DRIVER_CONST_BUFFERS)
78 #define R600_MAX_HW_CONST_BUFFERS 16
79 
80 /* start driver buffers after user buffers */
81 #define R600_BUFFER_INFO_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS)
82 #define R600_UCP_SIZE (4*4*8)
83 #define R600_CS_BLOCK_GRID_SIZE (8 * 4)
84 #define R600_TCS_DEFAULT_LEVELS_SIZE (6 * 4)
85 #define R600_BUFFER_INFO_OFFSET (R600_UCP_SIZE)
86 
87 /*
88  * We only access this buffer through vtx clauses hence it's fine to exist
89  * at index beyond 15.
90  */
91 #define R600_LDS_INFO_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 1)
92 /*
93  * Note GS doesn't use a constant buffer binding, just a resource index,
94  * so it's fine to have it exist at index beyond 15. I.e. it's not actually
95  * a const buffer, just a buffer resource.
96  */
97 #define R600_GS_RING_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 2)
98 /* Currently R600_MAX_CONST_BUFFERS just fits on the hw, which has a limit
99  * of 16 const buffers.
100  * UCP/SAMPLE_POSITIONS are never accessed by same shader stage so they can use the same id.
101  *
102  * In order to support d3d 11 mandated minimum of 15 user const buffers
103  * we'd have to squash all use cases into one driver buffer.
104  */
105 #define R600_MAX_CONST_BUFFER_SIZE (4096 * sizeof(float[4]))
106 
107 /* HW stages */
108 #define R600_HW_STAGE_PS 0
109 #define R600_HW_STAGE_VS 1
110 #define R600_HW_STAGE_GS 2
111 #define R600_HW_STAGE_ES 3
112 #define EG_HW_STAGE_LS 4
113 #define EG_HW_STAGE_HS 5
114 
115 #define R600_NUM_HW_STAGES 4
116 #define EG_NUM_HW_STAGES 6
117 
118 struct r600_context;
119 struct r600_bytecode;
120 union  r600_shader_key;
121 
122 /* This is an atom containing GPU commands that never change.
123  * This is supposed to be copied directly into the CS. */
124 struct r600_command_buffer {
125 	uint32_t *buf;
126 	unsigned num_dw;
127 	unsigned max_num_dw;
128 	unsigned pkt_flags;
129 };
130 
131 struct r600_db_state {
132 	struct r600_atom		atom;
133 	struct r600_surface		*rsurf;
134 };
135 
136 struct r600_db_misc_state {
137 	struct r600_atom		atom;
138 	bool				occlusion_queries_disabled;
139 	bool				flush_depthstencil_through_cb;
140 	bool				flush_depth_inplace;
141 	bool				flush_stencil_inplace;
142 	bool				copy_depth, copy_stencil;
143 	unsigned			copy_sample;
144 	unsigned			log_samples;
145 	unsigned			db_shader_control;
146 	bool				htile_clear;
147 	uint8_t				ps_conservative_z;
148 };
149 
150 struct r600_cb_misc_state {
151 	struct r600_atom atom;
152 	unsigned cb_color_control; /* this comes from blend state */
153 	unsigned blend_colormask; /* 8*4 bits for 8 RGBA colorbuffers */
154 	unsigned nr_cbufs;
155 	unsigned bound_cbufs_target_mask;
156 	unsigned nr_ps_color_outputs;
157 	unsigned ps_color_export_mask;
158 	unsigned image_rat_enabled_mask;
159 	unsigned buffer_rat_enabled_mask;
160 	bool multiwrite;
161 	bool dual_src_blend;
162 };
163 
164 struct r600_clip_misc_state {
165 	struct r600_atom atom;
166 	unsigned pa_cl_clip_cntl;   /* from rasterizer    */
167 	unsigned pa_cl_vs_out_cntl; /* from vertex shader */
168 	unsigned clip_plane_enable; /* from rasterizer    */
169 	unsigned cc_dist_mask;      /* from vertex shader */
170 	unsigned clip_dist_write;   /* from vertex shader */
171 	unsigned cull_dist_write;   /* from vertex shader */
172 	boolean clip_disable;       /* from vertex shader */
173 	boolean vs_out_viewport;    /* from vertex shader */
174 };
175 
176 struct r600_alphatest_state {
177 	struct r600_atom atom;
178 	unsigned sx_alpha_test_control; /* this comes from dsa state */
179 	unsigned sx_alpha_ref; /* this comes from dsa state */
180 	bool bypass;
181 	bool cb0_export_16bpc; /* from set_framebuffer_state */
182 };
183 
184 struct r600_vgt_state {
185 	struct r600_atom atom;
186 	uint32_t vgt_multi_prim_ib_reset_en;
187 	uint32_t vgt_multi_prim_ib_reset_indx;
188 	uint32_t vgt_indx_offset;
189 	bool last_draw_was_indirect;
190 };
191 
192 struct r600_blend_color {
193 	struct r600_atom atom;
194 	struct pipe_blend_color state;
195 };
196 
197 struct r600_clip_state {
198 	struct r600_atom atom;
199 	struct pipe_clip_state state;
200 };
201 
202 struct r600_cs_shader_state {
203 	struct r600_atom atom;
204 	unsigned kernel_index;
205 	unsigned pc;
206 	struct r600_pipe_compute *shader;
207 };
208 
209 struct r600_framebuffer {
210 	struct r600_atom atom;
211 	struct pipe_framebuffer_state state;
212 	unsigned compressed_cb_mask;
213 	unsigned nr_samples;
214 	bool export_16bpc;
215 	bool cb0_is_integer;
216 	bool is_msaa_resolve;
217 	bool dual_src_blend;
218 	bool do_update_surf_dirtiness;
219 };
220 
221 struct r600_sample_mask {
222 	struct r600_atom atom;
223 	uint16_t sample_mask; /* there are only 8 bits on EG, 16 bits on Cayman */
224 };
225 
226 struct r600_config_state {
227 	struct r600_atom atom;
228 	unsigned sq_gpr_resource_mgmt_1;
229 	unsigned sq_gpr_resource_mgmt_2;
230 	unsigned sq_gpr_resource_mgmt_3;
231 	bool dyn_gpr_enabled;
232 };
233 
234 struct r600_stencil_ref
235 {
236 	ubyte ref_value[2];
237 	ubyte valuemask[2];
238 	ubyte writemask[2];
239 };
240 
241 struct r600_stencil_ref_state {
242 	struct r600_atom atom;
243 	struct r600_stencil_ref state;
244 	struct pipe_stencil_ref pipe_state;
245 };
246 
247 struct r600_shader_stages_state {
248 	struct r600_atom atom;
249 	unsigned geom_enable;
250 };
251 
252 struct r600_gs_rings_state {
253 	struct r600_atom atom;
254 	unsigned enable;
255 	struct pipe_constant_buffer esgs_ring;
256 	struct pipe_constant_buffer gsvs_ring;
257 };
258 
259 /* This must start from 16. */
260 /* features */
261 #define DBG_NO_CP_DMA		(1 << 30)
262 /* shader backend */
263 #define DBG_NO_SB		(1 << 21)
264 #define DBG_SB_CS		(1 << 22)
265 #define DBG_SB_DRY_RUN	(1 << 23)
266 #define DBG_SB_STAT		(1 << 24)
267 #define DBG_SB_DUMP		(1 << 25)
268 #define DBG_SB_NO_FALLBACK	(1 << 26)
269 #define DBG_SB_DISASM	(1 << 27)
270 #define DBG_SB_SAFEMATH	(1 << 28)
271 
272 struct r600_screen {
273 	struct r600_common_screen	b;
274 	bool				has_msaa;
275 	bool				has_compressed_msaa_texturing;
276 	bool				has_atomics;
277 
278 	/*for compute global memory binding, we allocate stuff here, instead of
279 	 * buffers.
280 	 * XXX: Not sure if this is the best place for global_pool.  Also,
281 	 * it's not thread safe, so it won't work with multiple contexts. */
282 	struct compute_memory_pool *global_pool;
283 };
284 
285 struct r600_pipe_sampler_view {
286 	struct pipe_sampler_view	base;
287 	struct list_head		list;
288 	struct r600_resource		*tex_resource;
289 	uint32_t			tex_resource_words[8];
290 	bool				skip_mip_address_reloc;
291 	bool				is_stencil_sampler;
292 };
293 
294 struct r600_rasterizer_state {
295 	struct r600_command_buffer	buffer;
296 	boolean				flatshade;
297 	boolean				two_side;
298 	unsigned			sprite_coord_enable;
299 	unsigned                        clip_plane_enable;
300 	unsigned			pa_sc_line_stipple;
301 	unsigned			pa_cl_clip_cntl;
302 	unsigned			pa_su_sc_mode_cntl;
303 	float				offset_units;
304 	float				offset_scale;
305 	bool				offset_enable;
306 	bool				offset_units_unscaled;
307 	bool				scissor_enable;
308 	bool				multisample_enable;
309 	bool				clip_halfz;
310 	bool				rasterizer_discard;
311 };
312 
313 struct r600_poly_offset_state {
314 	struct r600_atom		atom;
315 	enum pipe_format		zs_format;
316 	float				offset_units;
317 	float				offset_scale;
318 	bool				offset_units_unscaled;
319 };
320 
321 struct r600_blend_state {
322 	struct r600_command_buffer	buffer;
323 	struct r600_command_buffer	buffer_no_blend;
324 	unsigned			cb_target_mask;
325 	unsigned			cb_color_control;
326 	unsigned			cb_color_control_no_blend;
327 	bool				dual_src_blend;
328 	bool				alpha_to_one;
329 };
330 
331 struct r600_dsa_state {
332 	struct r600_command_buffer	buffer;
333 	unsigned			alpha_ref;
334 	ubyte				valuemask[2];
335 	ubyte				writemask[2];
336 	unsigned			zwritemask;
337 	unsigned			sx_alpha_test_control;
338 };
339 
340 struct r600_pipe_shader;
341 
342 struct r600_pipe_shader_selector {
343 	struct r600_pipe_shader *current;
344 
345 	struct tgsi_token       *tokens;
346         struct nir_shader       *nir;
347 	struct pipe_stream_output_info  so;
348 	struct tgsi_shader_info		info;
349 
350 	unsigned	num_shaders;
351 
352 	enum pipe_shader_type	type;
353         enum pipe_shader_ir ir_type;
354 
355 	/* geometry shader properties */
356 	enum pipe_prim_type	gs_output_prim;
357 	unsigned		gs_max_out_vertices;
358 	unsigned		gs_num_invocations;
359 
360 	/* TCS/VS */
361 	uint64_t        lds_patch_outputs_written_mask;
362 	uint64_t        lds_outputs_written_mask;
363 	unsigned	nr_ps_max_color_exports;
364 };
365 
366 struct r600_pipe_sampler_state {
367 	uint32_t			tex_sampler_words[3];
368 	union pipe_color_union		border_color;
369 	bool				border_color_use;
370 	bool				seamless_cube_map;
371 };
372 
373 /* needed for blitter save */
374 #define NUM_TEX_UNITS 16
375 
376 struct r600_seamless_cube_map {
377 	struct r600_atom		atom;
378 	bool				enabled;
379 };
380 
381 struct r600_samplerview_state {
382 	struct r600_atom		atom;
383 	struct r600_pipe_sampler_view	*views[NUM_TEX_UNITS];
384 	uint32_t			enabled_mask;
385 	uint32_t			dirty_mask;
386 	uint32_t			compressed_depthtex_mask; /* which textures are depth */
387 	uint32_t			compressed_colortex_mask;
388 	boolean				dirty_buffer_constants;
389 };
390 
391 struct r600_sampler_states {
392 	struct r600_atom		atom;
393 	struct r600_pipe_sampler_state	*states[NUM_TEX_UNITS];
394 	uint32_t			enabled_mask;
395 	uint32_t			dirty_mask;
396 	uint32_t			has_bordercolor_mask; /* which states contain the border color */
397 };
398 
399 struct r600_textures_info {
400 	struct r600_samplerview_state	views;
401 	struct r600_sampler_states	states;
402 	bool				is_array_sampler[NUM_TEX_UNITS];
403 };
404 
405 struct r600_shader_driver_constants_info {
406 	/* currently 128 bytes for UCP/samplepos + sampler buffer constants */
407 	uint32_t			*constants;
408 	uint32_t			alloc_size;
409 	bool				texture_const_dirty;
410 	bool				vs_ucp_dirty;
411 	bool				ps_sample_pos_dirty;
412 	bool                            cs_block_grid_size_dirty;
413 	bool				tcs_default_levels_dirty;
414 };
415 
416 struct r600_constbuf_state
417 {
418 	struct r600_atom		atom;
419 	struct pipe_constant_buffer	cb[PIPE_MAX_CONSTANT_BUFFERS];
420 	uint32_t			enabled_mask;
421 	uint32_t			dirty_mask;
422 };
423 
424 struct r600_vertexbuf_state
425 {
426 	struct r600_atom		atom;
427 	struct pipe_vertex_buffer	vb[PIPE_MAX_ATTRIBS];
428 	uint32_t			enabled_mask; /* non-NULL buffers */
429 	uint32_t			dirty_mask;
430 };
431 
432 /* CSO (constant state object, in other words, immutable state). */
433 struct r600_cso_state
434 {
435 	struct r600_atom atom;
436 	void *cso; /* e.g. r600_blend_state */
437 	struct r600_command_buffer *cb;
438 };
439 
440 struct r600_fetch_shader {
441 	struct r600_resource		*buffer;
442 	unsigned			offset;
443 };
444 
445 struct r600_shader_state {
446 	struct r600_atom		atom;
447 	struct r600_pipe_shader *shader;
448 };
449 
450 struct r600_atomic_buffer_state {
451 	struct pipe_shader_buffer buffer[EG_MAX_ATOMIC_BUFFERS];
452 };
453 
454 struct r600_image_view {
455 	struct pipe_image_view base;
456 	uint32_t cb_color_base;
457 	uint32_t cb_color_pitch;
458 	uint32_t cb_color_slice;
459 	uint32_t cb_color_view;
460 	uint32_t cb_color_info;
461 	uint32_t cb_color_attrib;
462 	uint32_t cb_color_dim;
463 	uint32_t cb_color_fmask;
464 	uint32_t cb_color_fmask_slice;
465 	uint32_t immed_resource_words[8];
466 	uint32_t resource_words[8];
467 	bool skip_mip_address_reloc;
468 	uint32_t buf_size;
469 };
470 
471 struct r600_image_state {
472 	struct r600_atom atom;
473 	uint32_t                        enabled_mask;
474 	uint32_t                        dirty_mask;
475 	uint32_t			compressed_depthtex_mask;
476 	uint32_t			compressed_colortex_mask;
477 	boolean				dirty_buffer_constants;
478 	struct r600_image_view views[R600_MAX_IMAGES];
479 };
480 
481 /* Used to spill shader temps */
482 struct r600_scratch_buffer {
483 	struct r600_resource		*buffer;
484 	boolean					dirty;
485 	unsigned				size;
486 	unsigned				item_size;
487 };
488 
489 struct r600_context {
490 	struct r600_common_context	b;
491 	struct r600_screen		*screen;
492 	struct blitter_context		*blitter;
493 	struct u_suballocator		*allocator_fetch_shader;
494 
495 	/* Hardware info. */
496 	boolean				has_vertex_cache;
497 	unsigned			default_gprs[EG_NUM_HW_STAGES];
498 	unsigned                        current_gprs[EG_NUM_HW_STAGES];
499 	unsigned			r6xx_num_clause_temp_gprs;
500 
501 	/* Miscellaneous state objects. */
502 	void				*custom_dsa_flush;
503 	void				*custom_blend_resolve;
504 	void				*custom_blend_decompress;
505 	void                            *custom_blend_fastclear;
506 	/* With rasterizer discard, there doesn't have to be a pixel shader.
507 	 * In that case, we bind this one: */
508 	void				*dummy_pixel_shader;
509 	/* These dummy CMASK and FMASK buffers are used to get around the R6xx hardware
510 	 * bug where valid CMASK and FMASK are required to be present to avoid
511 	 * a hardlock in certain operations but aren't actually used
512 	 * for anything useful. */
513 	struct r600_resource		*dummy_fmask;
514 	struct r600_resource		*dummy_cmask;
515 
516 	/* State binding slots are here. */
517 	struct r600_atom		*atoms[R600_NUM_ATOMS];
518 	/* Dirty atom bitmask for fast tests */
519 	uint64_t			dirty_atoms;
520 	/* States for CS initialization. */
521 	struct r600_command_buffer	start_cs_cmd; /* invariant state mostly */
522 	/** Compute specific registers initializations.  The start_cs_cmd atom
523 	 *  must be emitted before start_compute_cs_cmd. */
524 	struct r600_command_buffer      start_compute_cs_cmd;
525 	/* Register states. */
526 	struct r600_alphatest_state	alphatest_state;
527 	struct r600_cso_state		blend_state;
528 	struct r600_blend_color		blend_color;
529 	struct r600_cb_misc_state	cb_misc_state;
530 	struct r600_clip_misc_state	clip_misc_state;
531 	struct r600_clip_state		clip_state;
532 	struct r600_db_misc_state	db_misc_state;
533 	struct r600_db_state		db_state;
534 	struct r600_cso_state		dsa_state;
535 	struct r600_framebuffer		framebuffer;
536 	struct r600_poly_offset_state	poly_offset_state;
537 	struct r600_cso_state		rasterizer_state;
538 	struct r600_sample_mask		sample_mask;
539 	struct r600_seamless_cube_map	seamless_cube_map;
540 	struct r600_config_state	config_state;
541 	struct r600_stencil_ref_state	stencil_ref;
542 	struct r600_vgt_state		vgt_state;
543 	struct r600_atomic_buffer_state atomic_buffer_state;
544 	/* only have images on fragment shader */
545 	struct r600_image_state         fragment_images;
546 	struct r600_image_state         compute_images;
547 	struct r600_image_state         fragment_buffers;
548 	struct r600_image_state         compute_buffers;
549 	/* Shaders and shader resources. */
550 	struct r600_cso_state		vertex_fetch_shader;
551 	struct r600_shader_state        hw_shader_stages[EG_NUM_HW_STAGES];
552 	struct r600_cs_shader_state	cs_shader_state;
553 	struct r600_shader_stages_state shader_stages;
554 	struct r600_gs_rings_state	gs_rings;
555 	struct r600_constbuf_state	constbuf_state[PIPE_SHADER_TYPES];
556 	struct r600_textures_info	samplers[PIPE_SHADER_TYPES];
557 
558 	struct r600_shader_driver_constants_info driver_consts[PIPE_SHADER_TYPES];
559 
560 	/** Vertex buffers for fetch shaders */
561 	struct r600_vertexbuf_state	vertex_buffer_state;
562 	/** Vertex buffers for compute shaders */
563 	struct r600_vertexbuf_state	cs_vertex_buffer_state;
564 
565 	/* Additional context states. */
566 	unsigned			compute_cb_target_mask;
567 	struct r600_pipe_shader_selector *ps_shader;
568 	struct r600_pipe_shader_selector *vs_shader;
569 	struct r600_pipe_shader_selector *gs_shader;
570 
571 	struct r600_pipe_shader_selector *tcs_shader;
572 	struct r600_pipe_shader_selector *tes_shader;
573 
574 	struct r600_pipe_shader_selector *fixed_func_tcs_shader;
575 
576 	struct r600_rasterizer_state	*rasterizer;
577 	bool				alpha_to_one;
578 	bool				force_blend_disable;
579 	bool                            gs_tri_strip_adj_fix;
580 	boolean				dual_src_blend;
581 	unsigned			zwritemask;
582 	unsigned			ps_iter_samples;
583 
584 	/* The list of all texture buffer objects in this context.
585 	 * This list is walked when a buffer is invalidated/reallocated and
586 	 * the GPU addresses are updated. */
587 	struct list_head		texture_buffers;
588 
589 	/* Last draw state (-1 = unset). */
590 	enum pipe_prim_type		last_primitive_type; /* Last primitive type used in draw_vbo. */
591 	enum pipe_prim_type		current_rast_prim; /* primitive type after TES, GS */
592 	enum pipe_prim_type		last_rast_prim;
593 	unsigned			last_start_instance;
594 
595 	void				*sb_context;
596 	struct r600_isa		*isa;
597 	float sample_positions[4 * 16];
598 	float tess_state[8];
599 	uint32_t cs_block_grid_sizes[8]; /* 3 for grid + 1 pad, 3 for block  + 1 pad*/
600 	struct r600_pipe_shader_selector *last_ls;
601 	struct r600_pipe_shader_selector *last_tcs;
602 	unsigned last_num_tcs_input_cp;
603 	unsigned lds_alloc;
604 
605 	struct r600_scratch_buffer scratch_buffers[MAX2(R600_NUM_HW_STAGES, EG_NUM_HW_STAGES)];
606 
607 	/* Debug state. */
608 	bool			is_debug;
609 	struct radeon_saved_cs	last_gfx;
610 	struct r600_resource	*last_trace_buf;
611 	struct r600_resource	*trace_buf;
612 	unsigned		trace_id;
613 
614 	bool cmd_buf_is_compute;
615 	struct pipe_resource *append_fence;
616 	uint32_t append_fence_id;
617 };
618 
r600_emit_command_buffer(struct radeon_cmdbuf * cs,struct r600_command_buffer * cb)619 static inline void r600_emit_command_buffer(struct radeon_cmdbuf *cs,
620 					    struct r600_command_buffer *cb)
621 {
622 	assert(cs->current.cdw + cb->num_dw <= cs->current.max_dw);
623 	memcpy(cs->current.buf + cs->current.cdw, cb->buf, 4 * cb->num_dw);
624 	cs->current.cdw += cb->num_dw;
625 }
626 
r600_set_atom_dirty(struct r600_context * rctx,struct r600_atom * atom,bool dirty)627 static inline void r600_set_atom_dirty(struct r600_context *rctx,
628 				       struct r600_atom *atom,
629 				       bool dirty)
630 {
631 	uint64_t mask;
632 
633 	assert(atom->id != 0);
634 	assert(atom->id < sizeof(mask) * 8);
635 	mask = 1ull << atom->id;
636 	if (dirty)
637 		rctx->dirty_atoms |= mask;
638 	else
639 		rctx->dirty_atoms &= ~mask;
640 }
641 
r600_mark_atom_dirty(struct r600_context * rctx,struct r600_atom * atom)642 static inline void r600_mark_atom_dirty(struct r600_context *rctx,
643 					struct r600_atom *atom)
644 {
645 	r600_set_atom_dirty(rctx, atom, true);
646 }
647 
r600_emit_atom(struct r600_context * rctx,struct r600_atom * atom)648 static inline void r600_emit_atom(struct r600_context *rctx, struct r600_atom *atom)
649 {
650 	atom->emit(&rctx->b, atom);
651 	r600_set_atom_dirty(rctx, atom, false);
652 }
653 
r600_set_cso_state(struct r600_context * rctx,struct r600_cso_state * state,void * cso)654 static inline void r600_set_cso_state(struct r600_context *rctx,
655 				      struct r600_cso_state *state, void *cso)
656 {
657 	state->cso = cso;
658 	r600_set_atom_dirty(rctx, &state->atom, cso != NULL);
659 }
660 
r600_set_cso_state_with_cb(struct r600_context * rctx,struct r600_cso_state * state,void * cso,struct r600_command_buffer * cb)661 static inline void r600_set_cso_state_with_cb(struct r600_context *rctx,
662 					      struct r600_cso_state *state, void *cso,
663 					      struct r600_command_buffer *cb)
664 {
665 	state->cb = cb;
666 	state->atom.num_dw = cb ? cb->num_dw : 0;
667 	r600_set_cso_state(rctx, state, cso);
668 }
669 
670 /* compute_memory_pool.c */
671 struct compute_memory_pool;
672 void compute_memory_pool_delete(struct compute_memory_pool* pool);
673 struct compute_memory_pool* compute_memory_pool_new(
674 	struct r600_screen *rscreen);
675 
676 /* evergreen_state.c */
677 struct pipe_sampler_view *
678 evergreen_create_sampler_view_custom(struct pipe_context *ctx,
679 				     struct pipe_resource *texture,
680 				     const struct pipe_sampler_view *state,
681 				     unsigned width0, unsigned height0,
682 				     unsigned force_level);
683 void evergreen_init_common_regs(struct r600_context *ctx,
684 				struct r600_command_buffer *cb,
685 				enum chip_class ctx_chip_class,
686 				enum radeon_family ctx_family,
687 				int ctx_drm_minor);
688 void cayman_init_common_regs(struct r600_command_buffer *cb,
689 			     enum chip_class ctx_chip_class,
690 			     enum radeon_family ctx_family,
691 			     int ctx_drm_minor);
692 
693 void evergreen_init_state_functions(struct r600_context *rctx);
694 void evergreen_init_atom_start_cs(struct r600_context *rctx);
695 void evergreen_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
696 void evergreen_update_es_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
697 void evergreen_update_gs_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
698 void evergreen_update_vs_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
699 void evergreen_update_ls_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
700 void evergreen_update_hs_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
701 void *evergreen_create_db_flush_dsa(struct r600_context *rctx);
702 void *evergreen_create_resolve_blend(struct r600_context *rctx);
703 void *evergreen_create_decompress_blend(struct r600_context *rctx);
704 void *evergreen_create_fastclear_blend(struct r600_context *rctx);
705 bool evergreen_is_format_supported(struct pipe_screen *screen,
706 				   enum pipe_format format,
707 				   enum pipe_texture_target target,
708 				   unsigned sample_count,
709 				   unsigned storage_sample_count,
710 				   unsigned usage);
711 void evergreen_init_color_surface(struct r600_context *rctx,
712 				  struct r600_surface *surf);
713 void evergreen_init_color_surface_rat(struct r600_context *rctx,
714 					struct r600_surface *surf);
715 void evergreen_update_db_shader_control(struct r600_context * rctx);
716 bool evergreen_adjust_gprs(struct r600_context *rctx);
717 void evergreen_setup_scratch_buffers(struct r600_context *rctx);
718 uint32_t evergreen_construct_rat_mask(struct r600_context *rctx, struct r600_cb_misc_state *a,
719 				      unsigned nr_cbufs);
720 /* r600_blit.c */
721 void r600_init_blit_functions(struct r600_context *rctx);
722 void r600_decompress_depth_textures(struct r600_context *rctx,
723 				    struct r600_samplerview_state *textures);
724 void r600_decompress_depth_images(struct r600_context *rctx,
725 				  struct r600_image_state *images);
726 void r600_decompress_color_textures(struct r600_context *rctx,
727 				    struct r600_samplerview_state *textures);
728 void r600_decompress_color_images(struct r600_context *rctx,
729 				  struct r600_image_state *images);
730 void r600_resource_copy_region(struct pipe_context *ctx,
731 			       struct pipe_resource *dst,
732 			       unsigned dst_level,
733 			       unsigned dstx, unsigned dsty, unsigned dstz,
734 			       struct pipe_resource *src,
735 			       unsigned src_level,
736 			       const struct pipe_box *src_box);
737 
738 /* r600_shader.c */
739 int r600_pipe_shader_create(struct pipe_context *ctx,
740 			    struct r600_pipe_shader *shader,
741 			    union r600_shader_key key);
742 
743 void r600_pipe_shader_destroy(struct pipe_context *ctx, struct r600_pipe_shader *shader);
744 
745 /* r600_state.c */
746 struct pipe_sampler_view *
747 r600_create_sampler_view_custom(struct pipe_context *ctx,
748 				struct pipe_resource *texture,
749 				const struct pipe_sampler_view *state,
750 				unsigned width_first_level, unsigned height_first_level);
751 void r600_init_state_functions(struct r600_context *rctx);
752 void r600_init_atom_start_cs(struct r600_context *rctx);
753 void r600_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
754 void r600_update_es_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
755 void r600_update_gs_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
756 void r600_update_vs_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
757 void *r600_create_db_flush_dsa(struct r600_context *rctx);
758 void *r600_create_resolve_blend(struct r600_context *rctx);
759 void *r700_create_resolve_blend(struct r600_context *rctx);
760 void *r600_create_decompress_blend(struct r600_context *rctx);
761 bool r600_adjust_gprs(struct r600_context *rctx);
762 bool r600_is_format_supported(struct pipe_screen *screen,
763 			      enum pipe_format format,
764 			      enum pipe_texture_target target,
765 			      unsigned sample_count,
766 			      unsigned storage_sample_count,
767 			      unsigned usage);
768 void r600_update_db_shader_control(struct r600_context * rctx);
769 void r600_setup_scratch_buffers(struct r600_context *rctx);
770 
771 /* r600_hw_context.c */
772 void r600_context_gfx_flush(void *context, unsigned flags,
773 			    struct pipe_fence_handle **fence);
774 void r600_begin_new_cs(struct r600_context *ctx);
775 void r600_flush_emit(struct r600_context *ctx);
776 void r600_need_cs_space(struct r600_context *ctx, unsigned num_dw, boolean count_draw_in, unsigned num_atomics);
777 void r600_emit_pfp_sync_me(struct r600_context *rctx);
778 void r600_cp_dma_copy_buffer(struct r600_context *rctx,
779 			     struct pipe_resource *dst, uint64_t dst_offset,
780 			     struct pipe_resource *src, uint64_t src_offset,
781 			     unsigned size);
782 void evergreen_cp_dma_clear_buffer(struct r600_context *rctx,
783 				   struct pipe_resource *dst, uint64_t offset,
784 				   unsigned size, uint32_t clear_value,
785 				   enum r600_coherency coher);
786 void r600_dma_copy_buffer(struct r600_context *rctx,
787 			  struct pipe_resource *dst,
788 			  struct pipe_resource *src,
789 			  uint64_t dst_offset,
790 			  uint64_t src_offset,
791 			  uint64_t size);
792 
793 /*
794  * evergreen_hw_context.c
795  */
796 void evergreen_dma_copy_buffer(struct r600_context *rctx,
797 			       struct pipe_resource *dst,
798 			       struct pipe_resource *src,
799 			       uint64_t dst_offset,
800 			       uint64_t src_offset,
801 			       uint64_t size);
802 void evergreen_setup_tess_constants(struct r600_context *rctx,
803 				    const struct pipe_draw_info *info,
804 				    unsigned *num_patches);
805 uint32_t evergreen_get_ls_hs_config(struct r600_context *rctx,
806 				    const struct pipe_draw_info *info,
807 				    unsigned num_patches);
808 void evergreen_set_ls_hs_config(struct r600_context *rctx,
809 				struct radeon_cmdbuf *cs,
810 				uint32_t ls_hs_config);
811 void evergreen_set_lds_alloc(struct r600_context *rctx,
812 			     struct radeon_cmdbuf *cs,
813 			     uint32_t lds_alloc);
814 
815 /* r600_state_common.c */
816 void r600_init_common_state_functions(struct r600_context *rctx);
817 void r600_emit_cso_state(struct r600_context *rctx, struct r600_atom *atom);
818 void r600_emit_alphatest_state(struct r600_context *rctx, struct r600_atom *atom);
819 void r600_emit_blend_color(struct r600_context *rctx, struct r600_atom *atom);
820 void r600_emit_vgt_state(struct r600_context *rctx, struct r600_atom *atom);
821 void r600_emit_clip_misc_state(struct r600_context *rctx, struct r600_atom *atom);
822 void r600_emit_stencil_ref(struct r600_context *rctx, struct r600_atom *atom);
823 void r600_emit_shader(struct r600_context *rctx, struct r600_atom *a);
824 void r600_add_atom(struct r600_context *rctx, struct r600_atom *atom, unsigned id);
825 void r600_init_atom(struct r600_context *rctx, struct r600_atom *atom, unsigned id,
826 		    void (*emit)(struct r600_context *ctx, struct r600_atom *state),
827 		    unsigned num_dw);
828 void r600_vertex_buffers_dirty(struct r600_context *rctx);
829 void r600_sampler_views_dirty(struct r600_context *rctx,
830 			      struct r600_samplerview_state *state);
831 void r600_sampler_states_dirty(struct r600_context *rctx,
832 			       struct r600_sampler_states *state);
833 void r600_constant_buffers_dirty(struct r600_context *rctx, struct r600_constbuf_state *state);
834 void r600_set_sample_locations_constant_buffer(struct r600_context *rctx);
835 void r600_setup_scratch_area_for_shader(struct r600_context *rctx,
836 	struct r600_pipe_shader *shader, struct r600_scratch_buffer *scratch,
837 	unsigned ring_base_reg, unsigned item_size_reg, unsigned ring_size_reg);
838 uint32_t r600_translate_stencil_op(int s_op);
839 uint32_t r600_translate_fill(uint32_t func);
840 unsigned r600_tex_wrap(unsigned wrap);
841 unsigned r600_tex_mipfilter(unsigned filter);
842 unsigned r600_tex_compare(unsigned compare);
843 bool sampler_state_needs_border_color(const struct pipe_sampler_state *state);
844 unsigned r600_get_swizzle_combined(const unsigned char *swizzle_format,
845 				   const unsigned char *swizzle_view,
846 				   boolean vtx);
847 uint32_t r600_translate_texformat(struct pipe_screen *screen, enum pipe_format format,
848 				  const unsigned char *swizzle_view,
849 				  uint32_t *word4_p, uint32_t *yuv_format_p,
850 				  bool do_endian_swap);
851 uint32_t r600_translate_colorformat(enum chip_class chip, enum pipe_format format,
852 				  bool do_endian_swap);
853 uint32_t r600_colorformat_endian_swap(uint32_t colorformat, bool do_endian_swap);
854 
855 /* r600_uvd.c */
856 struct pipe_video_codec *r600_uvd_create_decoder(struct pipe_context *context,
857 						   const struct pipe_video_codec *decoder);
858 
859 struct pipe_video_buffer *r600_video_buffer_create(struct pipe_context *pipe,
860 						   const struct pipe_video_buffer *tmpl);
861 
862 /*
863  * Helpers for building command buffers
864  */
865 
866 #define PKT3_SET_CONFIG_REG	0x68
867 #define PKT3_SET_CONTEXT_REG	0x69
868 #define PKT3_SET_CTL_CONST      0x6F
869 #define PKT3_SET_LOOP_CONST                    0x6C
870 
871 #define R600_CONFIG_REG_OFFSET	0x08000
872 #define R600_CONTEXT_REG_OFFSET 0x28000
873 #define R600_CTL_CONST_OFFSET   0x3CFF0
874 #define R600_LOOP_CONST_OFFSET                 0X0003E200
875 #define EG_LOOP_CONST_OFFSET               0x0003A200
876 
877 #define PKT_TYPE_S(x)                   (((unsigned)(x) & 0x3) << 30)
878 #define PKT_COUNT_S(x)                  (((unsigned)(x) & 0x3FFF) << 16)
879 #define PKT3_IT_OPCODE_S(x)             (((unsigned)(x) & 0xFF) << 8)
880 #define PKT3_PREDICATE(x)               (((x) >> 0) & 0x1)
881 #define PKT3(op, count, predicate) (PKT_TYPE_S(3) | PKT_COUNT_S(count) | PKT3_IT_OPCODE_S(op) | PKT3_PREDICATE(predicate))
882 
883 #define RADEON_CP_PACKET3_COMPUTE_MODE 0x00000002
884 
885 /*Evergreen Compute packet3*/
886 #define PKT3C(op, count, predicate) (PKT_TYPE_S(3) | PKT3_IT_OPCODE_S(op) | PKT_COUNT_S(count) | PKT3_PREDICATE(predicate) | RADEON_CP_PACKET3_COMPUTE_MODE)
887 
r600_store_value(struct r600_command_buffer * cb,unsigned value)888 static inline void r600_store_value(struct r600_command_buffer *cb, unsigned value)
889 {
890 	cb->buf[cb->num_dw++] = value;
891 }
892 
r600_store_array(struct r600_command_buffer * cb,unsigned num,unsigned * ptr)893 static inline void r600_store_array(struct r600_command_buffer *cb, unsigned num, unsigned *ptr)
894 {
895 	assert(cb->num_dw+num <= cb->max_num_dw);
896 	memcpy(&cb->buf[cb->num_dw], ptr, num * sizeof(ptr[0]));
897 	cb->num_dw += num;
898 }
899 
r600_store_config_reg_seq(struct r600_command_buffer * cb,unsigned reg,unsigned num)900 static inline void r600_store_config_reg_seq(struct r600_command_buffer *cb, unsigned reg, unsigned num)
901 {
902 	assert(reg < R600_CONTEXT_REG_OFFSET);
903 	assert(cb->num_dw+2+num <= cb->max_num_dw);
904 	cb->buf[cb->num_dw++] = PKT3(PKT3_SET_CONFIG_REG, num, 0);
905 	cb->buf[cb->num_dw++] = (reg - R600_CONFIG_REG_OFFSET) >> 2;
906 }
907 
908 /**
909  * Needs cb->pkt_flags set to  RADEON_CP_PACKET3_COMPUTE_MODE for compute
910  * shaders.
911  */
r600_store_context_reg_seq(struct r600_command_buffer * cb,unsigned reg,unsigned num)912 static inline void r600_store_context_reg_seq(struct r600_command_buffer *cb, unsigned reg, unsigned num)
913 {
914 	assert(reg >= R600_CONTEXT_REG_OFFSET && reg < R600_CTL_CONST_OFFSET);
915 	assert(cb->num_dw+2+num <= cb->max_num_dw);
916 	cb->buf[cb->num_dw++] = PKT3(PKT3_SET_CONTEXT_REG, num, 0) | cb->pkt_flags;
917 	cb->buf[cb->num_dw++] = (reg - R600_CONTEXT_REG_OFFSET) >> 2;
918 }
919 
920 /**
921  * Needs cb->pkt_flags set to  RADEON_CP_PACKET3_COMPUTE_MODE for compute
922  * shaders.
923  */
r600_store_ctl_const_seq(struct r600_command_buffer * cb,unsigned reg,unsigned num)924 static inline void r600_store_ctl_const_seq(struct r600_command_buffer *cb, unsigned reg, unsigned num)
925 {
926 	assert(reg >= R600_CTL_CONST_OFFSET);
927 	assert(cb->num_dw+2+num <= cb->max_num_dw);
928 	cb->buf[cb->num_dw++] = PKT3(PKT3_SET_CTL_CONST, num, 0) | cb->pkt_flags;
929 	cb->buf[cb->num_dw++] = (reg - R600_CTL_CONST_OFFSET) >> 2;
930 }
931 
r600_store_loop_const_seq(struct r600_command_buffer * cb,unsigned reg,unsigned num)932 static inline void r600_store_loop_const_seq(struct r600_command_buffer *cb, unsigned reg, unsigned num)
933 {
934 	assert(reg >= R600_LOOP_CONST_OFFSET);
935 	assert(cb->num_dw+2+num <= cb->max_num_dw);
936 	cb->buf[cb->num_dw++] = PKT3(PKT3_SET_LOOP_CONST, num, 0);
937 	cb->buf[cb->num_dw++] = (reg - R600_LOOP_CONST_OFFSET) >> 2;
938 }
939 
940 /**
941  * Needs cb->pkt_flags set to  RADEON_CP_PACKET3_COMPUTE_MODE for compute
942  * shaders.
943  */
eg_store_loop_const_seq(struct r600_command_buffer * cb,unsigned reg,unsigned num)944 static inline void eg_store_loop_const_seq(struct r600_command_buffer *cb, unsigned reg, unsigned num)
945 {
946 	assert(reg >= EG_LOOP_CONST_OFFSET);
947 	assert(cb->num_dw+2+num <= cb->max_num_dw);
948 	cb->buf[cb->num_dw++] = PKT3(PKT3_SET_LOOP_CONST, num, 0) | cb->pkt_flags;
949 	cb->buf[cb->num_dw++] = (reg - EG_LOOP_CONST_OFFSET) >> 2;
950 }
951 
r600_store_config_reg(struct r600_command_buffer * cb,unsigned reg,unsigned value)952 static inline void r600_store_config_reg(struct r600_command_buffer *cb, unsigned reg, unsigned value)
953 {
954 	r600_store_config_reg_seq(cb, reg, 1);
955 	r600_store_value(cb, value);
956 }
957 
r600_store_context_reg(struct r600_command_buffer * cb,unsigned reg,unsigned value)958 static inline void r600_store_context_reg(struct r600_command_buffer *cb, unsigned reg, unsigned value)
959 {
960 	r600_store_context_reg_seq(cb, reg, 1);
961 	r600_store_value(cb, value);
962 }
963 
r600_store_ctl_const(struct r600_command_buffer * cb,unsigned reg,unsigned value)964 static inline void r600_store_ctl_const(struct r600_command_buffer *cb, unsigned reg, unsigned value)
965 {
966 	r600_store_ctl_const_seq(cb, reg, 1);
967 	r600_store_value(cb, value);
968 }
969 
r600_store_loop_const(struct r600_command_buffer * cb,unsigned reg,unsigned value)970 static inline void r600_store_loop_const(struct r600_command_buffer *cb, unsigned reg, unsigned value)
971 {
972 	r600_store_loop_const_seq(cb, reg, 1);
973 	r600_store_value(cb, value);
974 }
975 
eg_store_loop_const(struct r600_command_buffer * cb,unsigned reg,unsigned value)976 static inline void eg_store_loop_const(struct r600_command_buffer *cb, unsigned reg, unsigned value)
977 {
978 	eg_store_loop_const_seq(cb, reg, 1);
979 	r600_store_value(cb, value);
980 }
981 
982 void r600_init_command_buffer(struct r600_command_buffer *cb, unsigned num_dw);
983 void r600_release_command_buffer(struct r600_command_buffer *cb);
984 
radeon_compute_set_context_reg_seq(struct radeon_cmdbuf * cs,unsigned reg,unsigned num)985 static inline void radeon_compute_set_context_reg_seq(struct radeon_cmdbuf *cs, unsigned reg, unsigned num)
986 {
987 	radeon_set_context_reg_seq(cs, reg, num);
988 	/* Set the compute bit on the packet header */
989 	cs->current.buf[cs->current.cdw - 2] |= RADEON_CP_PACKET3_COMPUTE_MODE;
990 }
991 
radeon_set_ctl_const_seq(struct radeon_cmdbuf * cs,unsigned reg,unsigned num)992 static inline void radeon_set_ctl_const_seq(struct radeon_cmdbuf *cs, unsigned reg, unsigned num)
993 {
994 	assert(reg >= R600_CTL_CONST_OFFSET);
995 	assert(cs->current.cdw + 2 + num <= cs->current.max_dw);
996 	radeon_emit(cs, PKT3(PKT3_SET_CTL_CONST, num, 0));
997 	radeon_emit(cs, (reg - R600_CTL_CONST_OFFSET) >> 2);
998 }
999 
radeon_compute_set_context_reg(struct radeon_cmdbuf * cs,unsigned reg,unsigned value)1000 static inline void radeon_compute_set_context_reg(struct radeon_cmdbuf *cs, unsigned reg, unsigned value)
1001 {
1002 	radeon_compute_set_context_reg_seq(cs, reg, 1);
1003 	radeon_emit(cs, value);
1004 }
1005 
radeon_set_context_reg_flag(struct radeon_cmdbuf * cs,unsigned reg,unsigned value,unsigned flag)1006 static inline void radeon_set_context_reg_flag(struct radeon_cmdbuf *cs, unsigned reg, unsigned value, unsigned flag)
1007 {
1008 	if (flag & RADEON_CP_PACKET3_COMPUTE_MODE) {
1009 		radeon_compute_set_context_reg(cs, reg, value);
1010 	} else {
1011 		radeon_set_context_reg(cs, reg, value);
1012 	}
1013 }
1014 
radeon_set_ctl_const(struct radeon_cmdbuf * cs,unsigned reg,unsigned value)1015 static inline void radeon_set_ctl_const(struct radeon_cmdbuf *cs, unsigned reg, unsigned value)
1016 {
1017 	radeon_set_ctl_const_seq(cs, reg, 1);
1018 	radeon_emit(cs, value);
1019 }
1020 
1021 /*
1022  * common helpers
1023  */
1024 
1025 /* 12.4 fixed-point */
r600_pack_float_12p4(float x)1026 static inline unsigned r600_pack_float_12p4(float x)
1027 {
1028 	return x <= 0    ? 0 :
1029 	       x >= 4096 ? 0xffff : x * 16;
1030 }
1031 
r600_get_flush_flags(enum r600_coherency coher)1032 static inline unsigned r600_get_flush_flags(enum r600_coherency coher)
1033 {
1034 	switch (coher) {
1035 	default:
1036 	case R600_COHERENCY_NONE:
1037 		return 0;
1038 	case R600_COHERENCY_SHADER:
1039 		return R600_CONTEXT_INV_CONST_CACHE |
1040 		       R600_CONTEXT_INV_VERTEX_CACHE |
1041 		       R600_CONTEXT_INV_TEX_CACHE |
1042 		       R600_CONTEXT_STREAMOUT_FLUSH;
1043 	case R600_COHERENCY_CB_META:
1044 		return R600_CONTEXT_FLUSH_AND_INV_CB |
1045 		       R600_CONTEXT_FLUSH_AND_INV_CB_META;
1046 	}
1047 }
1048 
1049 #define     V_028A6C_OUTPRIM_TYPE_POINTLIST            0
1050 #define     V_028A6C_OUTPRIM_TYPE_LINESTRIP            1
1051 #define     V_028A6C_OUTPRIM_TYPE_TRISTRIP             2
1052 
1053 unsigned r600_conv_prim_to_gs_out(unsigned mode);
1054 
1055 void eg_trace_emit(struct r600_context *rctx);
1056 void eg_dump_debug_state(struct pipe_context *ctx, FILE *f,
1057 			 unsigned flags);
1058 
1059 struct r600_pipe_shader_selector *r600_create_shader_state_tokens(struct pipe_context *ctx,
1060 								  const void *tokens,
1061 								  enum pipe_shader_ir,
1062 								  unsigned pipe_shader_type);
1063 int r600_shader_select(struct pipe_context *ctx,
1064 		       struct r600_pipe_shader_selector* sel,
1065 		       bool *dirty);
1066 
1067 void r600_delete_shader_selector(struct pipe_context *ctx,
1068 				 struct r600_pipe_shader_selector *sel);
1069 
1070 struct r600_shader_atomic;
1071 void evergreen_emit_atomic_buffer_setup_count(struct r600_context *rctx,
1072 					      struct r600_pipe_shader *cs_shader,
1073 					      struct r600_shader_atomic *combined_atomics,
1074 					      uint8_t *atomic_used_mask_p);
1075 void evergreen_emit_atomic_buffer_setup(struct r600_context *rctx,
1076 					bool is_compute,
1077 					struct r600_shader_atomic *combined_atomics,
1078 					uint8_t atomic_used_mask);
1079 void evergreen_emit_atomic_buffer_save(struct r600_context *rctx,
1080 				       bool is_compute,
1081 				       struct r600_shader_atomic *combined_atomics,
1082 				       uint8_t *atomic_used_mask_p);
1083 void r600_update_compressed_resource_state(struct r600_context *rctx, bool compute_only);
1084 
1085 void eg_setup_buffer_constants(struct r600_context *rctx, int shader_type);
1086 void r600_update_driver_const_buffers(struct r600_context *rctx, bool compute_only);
1087 #endif
1088