1/*
2 * Copyright 2012 Advanced Micro Devices, Inc.
3 * All Rights Reserved.
4 *
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the "Software"),
7 * to deal in the Software without restriction, including without limitation
8 * on the rights to use, copy, modify, merge, publish, distribute, sub
9 * license, and/or sell copies of the Software, and to permit persons to whom
10 * the Software is furnished to do so, subject to the following conditions:
11 *
12 * The above copyright notice and this permission notice (including the next
13 * paragraph) shall be included in all copies or substantial portions of the
14 * Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
20 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
21 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
22 * USE OR OTHER DEALINGS IN THE SOFTWARE.
23 */
24
25#ifndef SI_STATE_H
26#define SI_STATE_H
27
28#include "pipebuffer/pb_slab.h"
29#include "si_pm4.h"
30#include "util/u_blitter.h"
31
32#ifdef __cplusplus
33extern "C" {
34#endif
35
36#define SI_NUM_GRAPHICS_SHADERS (PIPE_SHADER_TESS_EVAL + 1)
37#define SI_NUM_SHADERS          (PIPE_SHADER_COMPUTE + 1)
38
39#define SI_NUM_VERTEX_BUFFERS SI_MAX_ATTRIBS
40#define SI_NUM_SAMPLERS       32 /* OpenGL textures units per shader */
41#define SI_NUM_CONST_BUFFERS  16
42#define SI_NUM_IMAGES         16
43#define SI_NUM_IMAGE_SLOTS    (SI_NUM_IMAGES * 2) /* the second half are FMASK slots */
44#define SI_NUM_SHADER_BUFFERS 32
45
46struct si_screen;
47struct si_shader;
48struct si_shader_ctx_state;
49struct si_shader_selector;
50struct si_texture;
51struct si_qbo_state;
52
53struct si_state_blend {
54   struct si_pm4_state pm4;
55   uint32_t cb_target_mask;
56   /* Set 0xf or 0x0 (4 bits) per render target if the following is
57    * true. ANDed with spi_shader_col_format.
58    */
59   unsigned cb_target_enabled_4bit;
60   unsigned blend_enable_4bit;
61   unsigned need_src_alpha_4bit;
62   unsigned commutative_4bit;
63   unsigned dcc_msaa_corruption_4bit;
64   bool alpha_to_coverage : 1;
65   bool alpha_to_one : 1;
66   bool dual_src_blend : 1;
67   bool logicop_enable : 1;
68   bool allows_noop_optimization : 1;
69};
70
71struct si_state_rasterizer {
72   struct si_pm4_state pm4;
73   /* poly offset states for 16-bit, 24-bit, and 32-bit zbuffers */
74   struct si_pm4_state *pm4_poly_offset;
75   unsigned pa_sc_line_stipple;
76   unsigned pa_cl_clip_cntl;
77   float line_width;
78   float max_point_size;
79   unsigned ngg_cull_flags : 8;
80   unsigned ngg_cull_flags_y_inverted : 8;
81   unsigned sprite_coord_enable : 8;
82   unsigned clip_plane_enable : 8;
83   unsigned half_pixel_center : 1;
84   unsigned flatshade : 1;
85   unsigned flatshade_first : 1;
86   unsigned two_side : 1;
87   unsigned multisample_enable : 1;
88   unsigned force_persample_interp : 1;
89   unsigned line_stipple_enable : 1;
90   unsigned poly_stipple_enable : 1;
91   unsigned line_smooth : 1;
92   unsigned poly_smooth : 1;
93   unsigned uses_poly_offset : 1;
94   unsigned clamp_fragment_color : 1;
95   unsigned clamp_vertex_color : 1;
96   unsigned rasterizer_discard : 1;
97   unsigned scissor_enable : 1;
98   unsigned clip_halfz : 1;
99   unsigned polygon_mode_is_lines : 1;
100   unsigned polygon_mode_is_points : 1;
101};
102
103struct si_dsa_stencil_ref_part {
104   uint8_t valuemask[2];
105   uint8_t writemask[2];
106};
107
108struct si_dsa_order_invariance {
109   /** Whether the final result in Z/S buffers is guaranteed to be
110    * invariant under changes to the order in which fragments arrive. */
111   bool zs : 1;
112
113   /** Whether the set of fragments that pass the combined Z/S test is
114    * guaranteed to be invariant under changes to the order in which
115    * fragments arrive. */
116   bool pass_set : 1;
117
118   /** Whether the last fragment that passes the combined Z/S test at each
119    * sample is guaranteed to be invariant under changes to the order in
120    * which fragments arrive. */
121   bool pass_last : 1;
122};
123
124struct si_state_dsa {
125   struct si_pm4_state pm4;
126   struct si_dsa_stencil_ref_part stencil_ref;
127
128   /* 0 = without stencil buffer, 1 = when both Z and S buffers are present */
129   struct si_dsa_order_invariance order_invariance[2];
130
131   ubyte alpha_func : 3;
132   bool depth_enabled : 1;
133   bool depth_write_enabled : 1;
134   bool stencil_enabled : 1;
135   bool stencil_write_enabled : 1;
136   bool db_can_write : 1;
137};
138
139struct si_stencil_ref {
140   struct pipe_stencil_ref state;
141   struct si_dsa_stencil_ref_part dsa_part;
142};
143
144struct si_vertex_elements {
145   struct si_resource *instance_divisor_factor_buffer;
146   uint32_t rsrc_word3[SI_MAX_ATTRIBS];
147   uint16_t src_offset[SI_MAX_ATTRIBS];
148   uint8_t fix_fetch[SI_MAX_ATTRIBS];
149   uint8_t format_size[SI_MAX_ATTRIBS];
150   uint8_t vertex_buffer_index[SI_MAX_ATTRIBS];
151
152   /* Bitmask of elements that always need a fixup to be applied. */
153   uint16_t fix_fetch_always;
154
155   /* Bitmask of elements whose fetch should always be opencoded. */
156   uint16_t fix_fetch_opencode;
157
158   /* Bitmask of elements which need to be opencoded if the vertex buffer
159    * is unaligned. */
160   uint16_t fix_fetch_unaligned;
161
162   /* For elements in fix_fetch_unaligned: whether the effective
163    * element load size as seen by the hardware is a dword (as opposed
164    * to a short).
165    */
166   uint16_t hw_load_is_dword;
167
168   /* Bitmask of vertex buffers requiring alignment check */
169   uint16_t vb_alignment_check_mask;
170
171   uint8_t count;
172
173   uint16_t first_vb_use_mask;
174   /* Vertex buffer descriptor list size aligned for optimal prefetch. */
175   uint16_t vb_desc_list_alloc_size;
176   uint16_t instance_divisor_is_one;     /* bitmask of inputs */
177   uint16_t instance_divisor_is_fetched; /* bitmask of inputs */
178};
179
180union si_state {
181   struct si_state_named {
182      struct si_state_blend *blend;
183      struct si_state_rasterizer *rasterizer;
184      struct si_state_dsa *dsa;
185      struct si_pm4_state *poly_offset;
186      struct si_shader *ls;
187      struct si_shader *hs;
188      struct si_shader *es;
189      struct si_shader *gs;
190      struct si_pm4_state *vgt_shader_config;
191      struct si_shader *vs;
192      struct si_shader *ps;
193   } named;
194   struct si_pm4_state *array[sizeof(struct si_state_named) / sizeof(struct si_pm4_state *)];
195};
196
197#define SI_STATE_IDX(name) (offsetof(union si_state, named.name) / sizeof(struct si_pm4_state *))
198#define SI_STATE_BIT(name) (1 << SI_STATE_IDX(name))
199#define SI_NUM_STATES      (sizeof(union si_state) / sizeof(struct si_pm4_state *))
200
201static inline unsigned si_states_that_always_roll_context(void)
202{
203   return (SI_STATE_BIT(blend) | SI_STATE_BIT(rasterizer) | SI_STATE_BIT(dsa) |
204           SI_STATE_BIT(poly_offset) | SI_STATE_BIT(vgt_shader_config));
205}
206
207union si_state_atoms {
208   struct si_atoms_s {
209      /* The order matters. */
210      struct si_atom render_cond;
211      struct si_atom streamout_begin;
212      struct si_atom streamout_enable; /* must be after streamout_begin */
213      struct si_atom framebuffer;
214      struct si_atom msaa_sample_locs;
215      struct si_atom db_render_state;
216      struct si_atom dpbb_state;
217      struct si_atom msaa_config;
218      struct si_atom sample_mask;
219      struct si_atom cb_render_state;
220      struct si_atom blend_color;
221      struct si_atom clip_regs;
222      struct si_atom clip_state;
223      struct si_atom shader_pointers;
224      struct si_atom guardband;
225      struct si_atom scissors;
226      struct si_atom viewports;
227      struct si_atom stencil_ref;
228      struct si_atom spi_map;
229      struct si_atom scratch_state;
230      struct si_atom window_rectangles;
231      struct si_atom shader_query;
232      struct si_atom ngg_cull_state;
233   } s;
234   struct si_atom array[sizeof(struct si_atoms_s) / sizeof(struct si_atom)];
235};
236
237#define SI_ATOM_BIT(name) (1 << (offsetof(union si_state_atoms, s.name) / sizeof(struct si_atom)))
238#define SI_NUM_ATOMS      (sizeof(union si_state_atoms) / sizeof(struct si_atom))
239
240static inline unsigned si_atoms_that_always_roll_context(void)
241{
242   return (SI_ATOM_BIT(streamout_begin) | SI_ATOM_BIT(streamout_enable) | SI_ATOM_BIT(framebuffer) |
243           SI_ATOM_BIT(msaa_sample_locs) | SI_ATOM_BIT(sample_mask) | SI_ATOM_BIT(blend_color) |
244           SI_ATOM_BIT(clip_state) | SI_ATOM_BIT(scissors) | SI_ATOM_BIT(viewports) |
245           SI_ATOM_BIT(stencil_ref) | SI_ATOM_BIT(scratch_state) | SI_ATOM_BIT(window_rectangles));
246}
247
248struct si_shader_data {
249   uint32_t sh_base[SI_NUM_SHADERS];
250};
251
252/* The list of registers whose emitted values are remembered by si_context. */
253enum si_tracked_reg
254{
255   SI_TRACKED_DB_RENDER_CONTROL, /* 2 consecutive registers */
256   SI_TRACKED_DB_COUNT_CONTROL,
257
258   SI_TRACKED_DB_RENDER_OVERRIDE2,
259   SI_TRACKED_DB_SHADER_CONTROL,
260
261   SI_TRACKED_CB_TARGET_MASK,
262   SI_TRACKED_CB_DCC_CONTROL,
263
264   SI_TRACKED_SX_PS_DOWNCONVERT, /* 3 consecutive registers */
265   SI_TRACKED_SX_BLEND_OPT_EPSILON,
266   SI_TRACKED_SX_BLEND_OPT_CONTROL,
267
268   SI_TRACKED_PA_SC_LINE_CNTL, /* 2 consecutive registers */
269   SI_TRACKED_PA_SC_AA_CONFIG,
270
271   SI_TRACKED_DB_EQAA,
272   SI_TRACKED_PA_SC_MODE_CNTL_1,
273
274   SI_TRACKED_PA_SU_PRIM_FILTER_CNTL,
275   SI_TRACKED_PA_SU_SMALL_PRIM_FILTER_CNTL,
276
277   SI_TRACKED_PA_CL_VS_OUT_CNTL,
278   SI_TRACKED_PA_CL_CLIP_CNTL,
279
280   SI_TRACKED_PA_SC_BINNER_CNTL_0,
281
282   SI_TRACKED_DB_VRS_OVERRIDE_CNTL,
283
284   SI_TRACKED_PA_CL_GB_VERT_CLIP_ADJ, /* 4 consecutive registers */
285   SI_TRACKED_PA_CL_GB_VERT_DISC_ADJ,
286   SI_TRACKED_PA_CL_GB_HORZ_CLIP_ADJ,
287   SI_TRACKED_PA_CL_GB_HORZ_DISC_ADJ,
288
289   SI_TRACKED_PA_SU_HARDWARE_SCREEN_OFFSET,
290   SI_TRACKED_PA_SU_VTX_CNTL,
291
292   SI_TRACKED_PA_SC_CLIPRECT_RULE,
293
294   SI_TRACKED_PA_SC_LINE_STIPPLE,
295
296   SI_TRACKED_VGT_ESGS_RING_ITEMSIZE,
297
298   SI_TRACKED_VGT_GSVS_RING_OFFSET_1, /* 3 consecutive registers */
299   SI_TRACKED_VGT_GSVS_RING_OFFSET_2,
300   SI_TRACKED_VGT_GSVS_RING_OFFSET_3,
301
302   SI_TRACKED_VGT_GSVS_RING_ITEMSIZE,
303   SI_TRACKED_VGT_GS_MAX_VERT_OUT,
304
305   SI_TRACKED_VGT_GS_VERT_ITEMSIZE, /* 4 consecutive registers */
306   SI_TRACKED_VGT_GS_VERT_ITEMSIZE_1,
307   SI_TRACKED_VGT_GS_VERT_ITEMSIZE_2,
308   SI_TRACKED_VGT_GS_VERT_ITEMSIZE_3,
309
310   SI_TRACKED_VGT_GS_INSTANCE_CNT,
311   SI_TRACKED_VGT_GS_ONCHIP_CNTL,
312   SI_TRACKED_VGT_GS_MAX_PRIMS_PER_SUBGROUP,
313   SI_TRACKED_VGT_GS_MODE,
314   SI_TRACKED_VGT_PRIMITIVEID_EN,
315   SI_TRACKED_VGT_REUSE_OFF,
316   SI_TRACKED_SPI_VS_OUT_CONFIG,
317   SI_TRACKED_PA_CL_VTE_CNTL,
318   SI_TRACKED_PA_CL_NGG_CNTL,
319   SI_TRACKED_GE_MAX_OUTPUT_PER_SUBGROUP,
320   SI_TRACKED_GE_NGG_SUBGRP_CNTL,
321
322   SI_TRACKED_SPI_SHADER_IDX_FORMAT, /* 2 consecutive registers */
323   SI_TRACKED_SPI_SHADER_POS_FORMAT,
324
325   SI_TRACKED_SPI_PS_INPUT_ENA, /* 2 consecutive registers */
326   SI_TRACKED_SPI_PS_INPUT_ADDR,
327
328   SI_TRACKED_SPI_BARYC_CNTL,
329   SI_TRACKED_SPI_PS_IN_CONTROL,
330
331   SI_TRACKED_SPI_SHADER_Z_FORMAT, /* 2 consecutive registers */
332   SI_TRACKED_SPI_SHADER_COL_FORMAT,
333
334   SI_TRACKED_CB_SHADER_MASK,
335   SI_TRACKED_VGT_TF_PARAM,
336   SI_TRACKED_VGT_VERTEX_REUSE_BLOCK_CNTL,
337
338   /* Non-context registers: */
339   SI_TRACKED_GE_PC_ALLOC,
340   SI_TRACKED_SPI_SHADER_PGM_RSRC3_GS,
341   SI_TRACKED_SPI_SHADER_PGM_RSRC4_GS,
342
343   SI_NUM_TRACKED_REGS,
344};
345
346struct si_tracked_regs {
347   uint64_t reg_saved;
348   uint32_t reg_value[SI_NUM_TRACKED_REGS];
349   uint32_t spi_ps_input_cntl[32];
350};
351
352/* Private read-write buffer slots. */
353enum
354{
355   SI_ES_RING_ESGS,
356   SI_GS_RING_ESGS,
357
358   SI_RING_GSVS,
359
360   SI_VS_STREAMOUT_BUF0,
361   SI_VS_STREAMOUT_BUF1,
362   SI_VS_STREAMOUT_BUF2,
363   SI_VS_STREAMOUT_BUF3,
364
365   SI_HS_CONST_DEFAULT_TESS_LEVELS,
366   SI_VS_CONST_INSTANCE_DIVISORS,
367   SI_VS_CONST_CLIP_PLANES,
368   SI_PS_CONST_POLY_STIPPLE,
369   SI_PS_CONST_SAMPLE_POSITIONS,
370
371   /* Image descriptor of color buffer 0 for KHR_blend_equation_advanced. */
372   SI_PS_IMAGE_COLORBUF0,
373   SI_PS_IMAGE_COLORBUF0_HI,
374   SI_PS_IMAGE_COLORBUF0_FMASK,
375   SI_PS_IMAGE_COLORBUF0_FMASK_HI,
376
377   GFX10_GS_QUERY_BUF,
378
379   SI_NUM_INTERNAL_BINDINGS,
380};
381
382/* Indices into sctx->descriptors, laid out so that gfx and compute pipelines
383 * are contiguous:
384 *
385 *  0 - rw buffers
386 *  1 - vertex const and shader buffers
387 *  2 - vertex samplers and images
388 *  3 - fragment const and shader buffer
389 *   ...
390 *  11 - compute const and shader buffers
391 *  12 - compute samplers and images
392 */
393enum
394{
395   SI_SHADER_DESCS_CONST_AND_SHADER_BUFFERS,
396   SI_SHADER_DESCS_SAMPLERS_AND_IMAGES,
397   SI_NUM_SHADER_DESCS,
398};
399
400#define SI_DESCS_INTERNAL      0
401#define SI_DESCS_FIRST_SHADER  1
402#define SI_DESCS_FIRST_COMPUTE (SI_DESCS_FIRST_SHADER + PIPE_SHADER_COMPUTE * SI_NUM_SHADER_DESCS)
403#define SI_NUM_DESCS           (SI_DESCS_FIRST_SHADER + SI_NUM_SHADERS * SI_NUM_SHADER_DESCS)
404
405#define SI_DESCS_SHADER_MASK(name)                                                                 \
406   u_bit_consecutive(SI_DESCS_FIRST_SHADER + PIPE_SHADER_##name * SI_NUM_SHADER_DESCS,             \
407                     SI_NUM_SHADER_DESCS)
408
409static inline unsigned si_const_and_shader_buffer_descriptors_idx(unsigned shader)
410{
411   return SI_DESCS_FIRST_SHADER + shader * SI_NUM_SHADER_DESCS +
412          SI_SHADER_DESCS_CONST_AND_SHADER_BUFFERS;
413}
414
415static inline unsigned si_sampler_and_image_descriptors_idx(unsigned shader)
416{
417   return SI_DESCS_FIRST_SHADER + shader * SI_NUM_SHADER_DESCS +
418          SI_SHADER_DESCS_SAMPLERS_AND_IMAGES;
419}
420
421/* This represents descriptors in memory, such as buffer resources,
422 * image resources, and sampler states.
423 */
424struct si_descriptors {
425   /* The list of descriptors in malloc'd memory. */
426   uint32_t *list;
427   /* The list in mapped GPU memory. */
428   uint32_t *gpu_list;
429
430   /* The buffer where the descriptors have been uploaded. */
431   struct si_resource *buffer;
432   uint64_t gpu_address;
433
434   /* The maximum number of descriptors. */
435   uint32_t num_elements;
436
437   /* Slots that are used by currently-bound shaders.
438    * It determines which slots are uploaded.
439    */
440   uint32_t first_active_slot;
441   uint32_t num_active_slots;
442
443   /* The SH register offset relative to USER_DATA*_0 where the pointer
444    * to the descriptor array will be stored. */
445   short shader_userdata_offset;
446   /* The size of one descriptor. */
447   ubyte element_dw_size;
448   /* If there is only one slot enabled, bind it directly instead of
449    * uploading descriptors. -1 if disabled. */
450   signed char slot_index_to_bind_directly;
451};
452
453struct si_buffer_resources {
454   struct pipe_resource **buffers; /* this has num_buffers elements */
455   unsigned *offsets;              /* this has num_buffers elements */
456
457   enum radeon_bo_priority priority : 6;
458   enum radeon_bo_priority priority_constbuf : 6;
459
460   /* The i-th bit is set if that element is enabled (non-NULL resource). */
461   uint64_t enabled_mask;
462   uint64_t writable_mask;
463};
464
465#define si_pm4_state_changed(sctx, member)                                                         \
466   ((sctx)->queued.named.member != (sctx)->emitted.named.member)
467
468#define si_pm4_state_enabled_and_changed(sctx, member)                                             \
469   ((sctx)->queued.named.member && si_pm4_state_changed(sctx, member))
470
471#define si_pm4_bind_state(sctx, member, value)                                                     \
472   do {                                                                                            \
473      (sctx)->queued.named.member = (value);                                                       \
474      if (value && value != (sctx)->emitted.named.member)                                          \
475         (sctx)->dirty_states |= SI_STATE_BIT(member);                                             \
476      else                                                                                         \
477         (sctx)->dirty_states &= ~SI_STATE_BIT(member);                                            \
478   } while (0)
479
480/* si_descriptors.c */
481void si_set_mutable_tex_desc_fields(struct si_screen *sscreen, struct si_texture *tex,
482                                    const struct legacy_surf_level *base_level_info,
483                                    unsigned base_level, unsigned first_level, unsigned block_width,
484                                    /* restrict decreases overhead of si_set_sampler_view_desc ~8x. */
485                                    bool is_stencil, uint16_t access, uint32_t * restrict state);
486void si_update_ps_colorbuf0_slot(struct si_context *sctx);
487void si_invalidate_inlinable_uniforms(struct si_context *sctx, enum pipe_shader_type shader);
488void si_get_pipe_constant_buffer(struct si_context *sctx, uint shader, uint slot,
489                                 struct pipe_constant_buffer *cbuf);
490void si_get_shader_buffers(struct si_context *sctx, enum pipe_shader_type shader, uint start_slot,
491                           uint count, struct pipe_shader_buffer *sbuf);
492void si_set_ring_buffer(struct si_context *sctx, uint slot, struct pipe_resource *buffer,
493                        unsigned stride, unsigned num_records, bool add_tid, bool swizzle,
494                        unsigned element_size, unsigned index_stride, uint64_t offset);
495void si_init_all_descriptors(struct si_context *sctx);
496bool si_upload_graphics_shader_descriptors(struct si_context *sctx);
497bool si_upload_compute_shader_descriptors(struct si_context *sctx);
498void si_release_all_descriptors(struct si_context *sctx);
499void si_gfx_resources_add_all_to_bo_list(struct si_context *sctx);
500void si_compute_resources_add_all_to_bo_list(struct si_context *sctx);
501bool si_gfx_resources_check_encrypted(struct si_context *sctx);
502bool si_compute_resources_check_encrypted(struct si_context *sctx);
503void si_shader_pointers_mark_dirty(struct si_context *sctx);
504void si_add_all_descriptors_to_bo_list(struct si_context *sctx);
505void si_update_all_texture_descriptors(struct si_context *sctx);
506void si_shader_change_notify(struct si_context *sctx);
507void si_update_needs_color_decompress_masks(struct si_context *sctx);
508void si_emit_graphics_shader_pointers(struct si_context *sctx);
509void si_emit_compute_shader_pointers(struct si_context *sctx);
510void si_set_internal_const_buffer(struct si_context *sctx, uint slot,
511                                  const struct pipe_constant_buffer *input);
512void si_set_internal_shader_buffer(struct si_context *sctx, uint slot,
513                                   const struct pipe_shader_buffer *sbuffer);
514void si_set_active_descriptors(struct si_context *sctx, unsigned desc_idx,
515                               uint64_t new_active_mask);
516void si_set_active_descriptors_for_shader(struct si_context *sctx, struct si_shader_selector *sel);
517bool si_bindless_descriptor_can_reclaim_slab(void *priv, struct pb_slab_entry *entry);
518struct pb_slab *si_bindless_descriptor_slab_alloc(void *priv, unsigned heap, unsigned entry_size,
519                                                  unsigned group_index);
520void si_bindless_descriptor_slab_free(void *priv, struct pb_slab *pslab);
521void si_rebind_buffer(struct si_context *sctx, struct pipe_resource *buf);
522/* si_state.c */
523uint32_t si_translate_colorformat(enum chip_class chip_class, enum pipe_format format);
524void si_init_state_compute_functions(struct si_context *sctx);
525void si_init_state_functions(struct si_context *sctx);
526void si_init_screen_state_functions(struct si_screen *sscreen);
527void si_init_cs_preamble_state(struct si_context *sctx, bool uses_reg_shadowing);
528void si_make_buffer_descriptor(struct si_screen *screen, struct si_resource *buf,
529                               enum pipe_format format, unsigned offset, unsigned size,
530                               uint32_t *state);
531struct pipe_sampler_view *si_create_sampler_view_custom(struct pipe_context *ctx,
532                                                        struct pipe_resource *texture,
533                                                        const struct pipe_sampler_view *state,
534                                                        unsigned width0, unsigned height0,
535                                                        unsigned force_level);
536void si_update_fb_dirtiness_after_rendering(struct si_context *sctx);
537void si_mark_display_dcc_dirty(struct si_context *sctx, struct si_texture *tex);
538void si_update_ps_iter_samples(struct si_context *sctx);
539void si_save_qbo_state(struct si_context *sctx, struct si_qbo_state *st);
540void si_restore_qbo_state(struct si_context *sctx, struct si_qbo_state *st);
541void si_set_occlusion_query_state(struct si_context *sctx, bool old_perfect_enable);
542
543struct si_fast_udiv_info32 {
544   unsigned multiplier; /* the "magic number" multiplier */
545   unsigned pre_shift;  /* shift for the dividend before multiplying */
546   unsigned post_shift; /* shift for the dividend after multiplying */
547   int increment;       /* 0 or 1; if set then increment the numerator, using one of
548                           the two strategies */
549};
550
551struct si_fast_udiv_info32 si_compute_fast_udiv_info32(uint32_t D, unsigned num_bits);
552
553/* si_state_binning.c */
554void si_emit_dpbb_state(struct si_context *sctx);
555
556/* si_state_shaders.c */
557void si_get_ir_cache_key(struct si_shader_selector *sel, bool ngg, bool es,
558                         unsigned char ir_sha1_cache_key[20]);
559bool si_shader_cache_load_shader(struct si_screen *sscreen, unsigned char ir_sha1_cache_key[20],
560                                 struct si_shader *shader);
561void si_shader_cache_insert_shader(struct si_screen *sscreen, unsigned char ir_sha1_cache_key[20],
562                                   struct si_shader *shader, bool insert_into_disk_cache);
563bool si_shader_mem_ordered(struct si_shader *shader);
564void si_init_screen_live_shader_cache(struct si_screen *sscreen);
565void si_init_shader_functions(struct si_context *sctx);
566bool si_init_shader_cache(struct si_screen *sscreen);
567void si_destroy_shader_cache(struct si_screen *sscreen);
568void si_schedule_initial_compile(struct si_context *sctx, gl_shader_stage stage,
569                                 struct util_queue_fence *ready_fence,
570                                 struct si_compiler_ctx_state *compiler_ctx_state, void *job,
571                                 util_queue_execute_func execute);
572void si_get_active_slot_masks(const struct si_shader_info *info, uint64_t *const_and_shader_buffers,
573                              uint64_t *samplers_and_images);
574int si_shader_select_with_key(struct si_context *sctx, struct si_shader_ctx_state *state,
575                              const struct si_shader_key *key, int thread_index,
576                              bool optimized_or_none);
577int si_shader_select(struct pipe_context *ctx, struct si_shader_ctx_state *state);
578void si_vs_key_update_inputs(struct si_context *sctx);
579void si_get_vs_key_inputs(struct si_context *sctx, struct si_shader_key *key,
580                          struct si_vs_prolog_bits *prolog_key);
581void si_update_ps_inputs_read_or_disabled(struct si_context *sctx);
582void si_update_ps_kill_enable(struct si_context *sctx);
583void si_update_vrs_flat_shading(struct si_context *sctx);
584unsigned si_get_input_prim(const struct si_shader_selector *gs, const struct si_shader_key *key);
585bool si_update_ngg(struct si_context *sctx);
586void si_ps_key_update_framebuffer(struct si_context *sctx);
587void si_ps_key_update_framebuffer_blend(struct si_context *sctx);
588void si_ps_key_update_blend_rasterizer(struct si_context *sctx);
589void si_ps_key_update_rasterizer(struct si_context *sctx);
590void si_ps_key_update_dsa(struct si_context *sctx);
591void si_ps_key_update_sample_shading(struct si_context *sctx);
592void si_ps_key_update_framebuffer_rasterizer_sample_shading(struct si_context *sctx);
593void si_init_tess_factor_ring(struct si_context *sctx);
594bool si_update_gs_ring_buffers(struct si_context *sctx);
595bool si_update_spi_tmpring_size(struct si_context *sctx, unsigned bytes);
596
597/* si_state_draw.cpp */
598void si_set_vertex_buffer_descriptor(struct si_screen *sscreen, struct si_vertex_elements *velems,
599                                     struct pipe_vertex_buffer *vb, unsigned element_index,
600                                     uint32_t *out);
601void si_init_draw_functions_GFX6(struct si_context *sctx);
602void si_init_draw_functions_GFX7(struct si_context *sctx);
603void si_init_draw_functions_GFX8(struct si_context *sctx);
604void si_init_draw_functions_GFX9(struct si_context *sctx);
605void si_init_draw_functions_GFX10(struct si_context *sctx);
606void si_init_draw_functions_GFX10_3(struct si_context *sctx);
607void si_init_spi_map_functions(struct si_context *sctx);
608
609/* si_state_msaa.c */
610void si_init_msaa_functions(struct si_context *sctx);
611void si_emit_sample_locations(struct radeon_cmdbuf *cs, int nr_samples);
612
613/* si_state_streamout.c */
614void si_streamout_buffers_dirty(struct si_context *sctx);
615void si_emit_streamout_end(struct si_context *sctx);
616void si_update_prims_generated_query_state(struct si_context *sctx, unsigned type, int diff);
617void si_init_streamout_functions(struct si_context *sctx);
618
619static inline unsigned si_get_constbuf_slot(unsigned slot)
620{
621   /* Constant buffers are in slots [32..47], ascending */
622   return SI_NUM_SHADER_BUFFERS + slot;
623}
624
625static inline unsigned si_get_shaderbuf_slot(unsigned slot)
626{
627   /* shader buffers are in slots [31..0], descending */
628   return SI_NUM_SHADER_BUFFERS - 1 - slot;
629}
630
631static inline unsigned si_get_sampler_slot(unsigned slot)
632{
633   /* 32 samplers are in sampler slots [16..47], 16 dw per slot, ascending */
634   /* those are equivalent to image slots [32..95], 8 dw per slot, ascending  */
635   return SI_NUM_IMAGE_SLOTS / 2 + slot;
636}
637
638static inline unsigned si_get_image_slot(unsigned slot)
639{
640   /* image slots are in [31..0] (sampler slots [15..0]), descending */
641   /* images are in slots [31..16], while FMASKs are in slots [15..0] */
642   return SI_NUM_IMAGE_SLOTS - 1 - slot;
643}
644
645#ifdef __cplusplus
646}
647#endif
648
649#endif
650