radeonsi: pack struct si_descriptors better
[mesa.git] / src / gallium / drivers / radeonsi / si_state.h
1 /*
2 * Copyright 2012 Advanced Micro Devices, Inc.
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 * Christian König <christian.koenig@amd.com>
25 */
26
27 #ifndef SI_STATE_H
28 #define SI_STATE_H
29
30 #include "si_pm4.h"
31 #include "radeon/r600_pipe_common.h"
32
33 #define SI_NUM_GRAPHICS_SHADERS (PIPE_SHADER_TESS_EVAL+1)
34 #define SI_NUM_SHADERS (PIPE_SHADER_COMPUTE+1)
35
36 #define SI_MAX_ATTRIBS 16
37 #define SI_NUM_VERTEX_BUFFERS SI_MAX_ATTRIBS
38 #define SI_NUM_SAMPLERS 32 /* OpenGL textures units per shader */
39 #define SI_NUM_CONST_BUFFERS 16
40 #define SI_NUM_IMAGES 16
41 #define SI_NUM_SHADER_BUFFERS 16
42
43 struct si_screen;
44 struct si_shader;
45 struct si_shader_selector;
46
47 struct si_state_blend {
48 struct si_pm4_state pm4;
49 uint32_t cb_target_mask;
50 bool alpha_to_coverage;
51 bool alpha_to_one;
52 bool dual_src_blend;
53 /* Set 0xf or 0x0 (4 bits) per render target if the following is
54 * true. ANDed with spi_shader_col_format.
55 */
56 unsigned blend_enable_4bit;
57 unsigned need_src_alpha_4bit;
58 };
59
60 struct si_state_rasterizer {
61 struct si_pm4_state pm4;
62 /* poly offset states for 16-bit, 24-bit, and 32-bit zbuffers */
63 struct si_pm4_state *pm4_poly_offset;
64 unsigned pa_sc_line_stipple;
65 unsigned pa_cl_clip_cntl;
66 unsigned sprite_coord_enable:8;
67 unsigned clip_plane_enable:8;
68 unsigned flatshade:1;
69 unsigned two_side:1;
70 unsigned multisample_enable:1;
71 unsigned force_persample_interp:1;
72 unsigned line_stipple_enable:1;
73 unsigned poly_stipple_enable:1;
74 unsigned line_smooth:1;
75 unsigned poly_smooth:1;
76 unsigned uses_poly_offset:1;
77 unsigned clamp_fragment_color:1;
78 unsigned clamp_vertex_color:1;
79 unsigned rasterizer_discard:1;
80 unsigned scissor_enable:1;
81 unsigned clip_halfz:1;
82 };
83
84 struct si_dsa_stencil_ref_part {
85 uint8_t valuemask[2];
86 uint8_t writemask[2];
87 };
88
89 struct si_state_dsa {
90 struct si_pm4_state pm4;
91 unsigned alpha_func;
92 struct si_dsa_stencil_ref_part stencil_ref;
93 };
94
95 struct si_stencil_ref {
96 struct r600_atom atom;
97 struct pipe_stencil_ref state;
98 struct si_dsa_stencil_ref_part dsa_part;
99 };
100
101 struct si_vertex_elements
102 {
103 uint32_t instance_divisors[SI_MAX_ATTRIBS];
104 uint32_t rsrc_word3[SI_MAX_ATTRIBS];
105 uint16_t src_offset[SI_MAX_ATTRIBS];
106 uint8_t fix_fetch[SI_MAX_ATTRIBS];
107 uint8_t format_size[SI_MAX_ATTRIBS];
108 uint8_t vertex_buffer_index[SI_MAX_ATTRIBS];
109
110 uint8_t count;
111 bool uses_instance_divisors;
112
113 uint16_t first_vb_use_mask;
114 /* Vertex buffer descriptor list size aligned for optimal prefetch. */
115 uint16_t desc_list_byte_size;
116 };
117
118 union si_state {
119 struct {
120 struct si_state_blend *blend;
121 struct si_state_rasterizer *rasterizer;
122 struct si_state_dsa *dsa;
123 struct si_pm4_state *poly_offset;
124 struct si_pm4_state *ls;
125 struct si_pm4_state *hs;
126 struct si_pm4_state *es;
127 struct si_pm4_state *gs;
128 struct si_pm4_state *vgt_shader_config;
129 struct si_pm4_state *vs;
130 struct si_pm4_state *ps;
131 } named;
132 struct si_pm4_state *array[0];
133 };
134
135 #define SI_NUM_STATES (sizeof(union si_state) / sizeof(struct si_pm4_state *))
136
137 union si_state_atoms {
138 struct {
139 /* The order matters. */
140 struct r600_atom *prefetch_L2;
141 struct r600_atom *render_cond;
142 struct r600_atom *streamout_begin;
143 struct r600_atom *streamout_enable; /* must be after streamout_begin */
144 struct r600_atom *framebuffer;
145 struct r600_atom *msaa_sample_locs;
146 struct r600_atom *db_render_state;
147 struct r600_atom *msaa_config;
148 struct r600_atom *sample_mask;
149 struct r600_atom *cb_render_state;
150 struct r600_atom *blend_color;
151 struct r600_atom *clip_regs;
152 struct r600_atom *clip_state;
153 struct r600_atom *shader_userdata;
154 struct r600_atom *scissors;
155 struct r600_atom *viewports;
156 struct r600_atom *stencil_ref;
157 struct r600_atom *spi_map;
158 struct r600_atom *scratch_state;
159 } s;
160 struct r600_atom *array[0];
161 };
162
163 #define SI_NUM_ATOMS (sizeof(union si_state_atoms)/sizeof(struct r600_atom*))
164
165 struct si_shader_data {
166 struct r600_atom atom;
167 uint32_t sh_base[SI_NUM_SHADERS];
168 };
169
170 /* Private read-write buffer slots. */
171 enum {
172 SI_ES_RING_ESGS,
173 SI_GS_RING_ESGS,
174
175 SI_RING_GSVS,
176
177 SI_VS_STREAMOUT_BUF0,
178 SI_VS_STREAMOUT_BUF1,
179 SI_VS_STREAMOUT_BUF2,
180 SI_VS_STREAMOUT_BUF3,
181
182 SI_HS_CONST_DEFAULT_TESS_LEVELS,
183 SI_VS_CONST_CLIP_PLANES,
184 SI_PS_CONST_POLY_STIPPLE,
185 SI_PS_CONST_SAMPLE_POSITIONS,
186
187 SI_NUM_RW_BUFFERS,
188 };
189
190 /* Indices into sctx->descriptors, laid out so that gfx and compute pipelines
191 * are contiguous:
192 *
193 * 0 - rw buffers
194 * 1 - vertex const buffers
195 * 2 - vertex shader buffers
196 * ...
197 * 5 - fragment const buffers
198 * ...
199 * 21 - compute const buffers
200 * ...
201 */
202 enum {
203 SI_SHADER_DESCS_CONST_AND_SHADER_BUFFERS,
204 SI_SHADER_DESCS_SAMPLERS_AND_IMAGES,
205 SI_NUM_SHADER_DESCS,
206 };
207
208 #define SI_DESCS_RW_BUFFERS 0
209 #define SI_DESCS_FIRST_SHADER 1
210 #define SI_DESCS_FIRST_COMPUTE (SI_DESCS_FIRST_SHADER + \
211 PIPE_SHADER_COMPUTE * SI_NUM_SHADER_DESCS)
212 #define SI_NUM_DESCS (SI_DESCS_FIRST_SHADER + \
213 SI_NUM_SHADERS * SI_NUM_SHADER_DESCS)
214
215 /* This represents descriptors in memory, such as buffer resources,
216 * image resources, and sampler states.
217 */
218 struct si_descriptors {
219 /* The list of descriptors in malloc'd memory. */
220 uint32_t *list;
221 /* The list in mapped GPU memory. */
222 uint32_t *gpu_list;
223 /* Slots that have been changed and need to be uploaded. */
224 uint64_t dirty_mask;
225
226 /* The buffer where the descriptors have been uploaded. */
227 struct r600_resource *buffer;
228 int buffer_offset; /* can be negative if not using lower slots */
229
230 /* The size of one descriptor. */
231 ubyte element_dw_size;
232 /* The maximum number of descriptors. */
233 ubyte num_elements;
234
235 /* Offset in CE RAM */
236 uint16_t ce_offset;
237
238 /* Slots allocated in CE RAM. If we get active slots outside of this
239 * range, direct uploads to memory will be used instead. This basically
240 * governs switching between onchip (CE) and offchip (upload) modes.
241 */
242 ubyte first_ce_slot;
243 ubyte num_ce_slots;
244
245 /* Slots that are used by currently-bound shaders.
246 * With CE: It determines which slots are dumped to L2.
247 * It doesn't skip uploads to CE RAM.
248 * Without CE: It determines which slots are uploaded.
249 */
250 ubyte first_active_slot;
251 ubyte num_active_slots;
252
253 /* Whether CE is used to upload this descriptor array. */
254 bool uses_ce;
255
256 /* The SGPR index where the 64-bit pointer to the descriptor array will
257 * be stored. */
258 ubyte shader_userdata_offset;
259 };
260
261 struct si_sampler_views {
262 struct pipe_sampler_view *views[SI_NUM_SAMPLERS];
263 struct si_sampler_state *sampler_states[SI_NUM_SAMPLERS];
264
265 /* The i-th bit is set if that element is enabled (non-NULL resource). */
266 unsigned enabled_mask;
267 };
268
269 struct si_buffer_resources {
270 enum radeon_bo_usage shader_usage; /* READ, WRITE, or READWRITE */
271 enum radeon_bo_usage shader_usage_constbuf;
272 enum radeon_bo_priority priority;
273 enum radeon_bo_priority priority_constbuf;
274 struct pipe_resource **buffers; /* this has num_buffers elements */
275
276 /* The i-th bit is set if that element is enabled (non-NULL resource). */
277 unsigned enabled_mask;
278 };
279
280 #define si_pm4_block_idx(member) \
281 (offsetof(union si_state, named.member) / sizeof(struct si_pm4_state *))
282
283 #define si_pm4_state_changed(sctx, member) \
284 ((sctx)->queued.named.member != (sctx)->emitted.named.member)
285
286 #define si_pm4_bind_state(sctx, member, value) \
287 do { \
288 (sctx)->queued.named.member = (value); \
289 (sctx)->dirty_states |= 1 << si_pm4_block_idx(member); \
290 } while(0)
291
292 #define si_pm4_delete_state(sctx, member, value) \
293 do { \
294 if ((sctx)->queued.named.member == (value)) { \
295 (sctx)->queued.named.member = NULL; \
296 } \
297 si_pm4_free_state(sctx, (struct si_pm4_state *)(value), \
298 si_pm4_block_idx(member)); \
299 } while(0)
300
301 /* si_descriptors.c */
302 void si_ce_save_all_descriptors_at_ib_end(struct si_context* sctx);
303 void si_ce_restore_all_descriptors_at_ib_start(struct si_context *sctx);
304 void si_ce_enable_loads(struct radeon_winsys_cs *ib);
305 void si_set_mutable_tex_desc_fields(struct si_screen *sscreen,
306 struct r600_texture *tex,
307 const struct legacy_surf_level *base_level_info,
308 unsigned base_level, unsigned first_level,
309 unsigned block_width, bool is_stencil,
310 uint32_t *state);
311 void si_get_pipe_constant_buffer(struct si_context *sctx, uint shader,
312 uint slot, struct pipe_constant_buffer *cbuf);
313 void si_get_shader_buffers(struct si_context *sctx,
314 enum pipe_shader_type shader,
315 uint start_slot, uint count,
316 struct pipe_shader_buffer *sbuf);
317 void si_set_ring_buffer(struct pipe_context *ctx, uint slot,
318 struct pipe_resource *buffer,
319 unsigned stride, unsigned num_records,
320 bool add_tid, bool swizzle,
321 unsigned element_size, unsigned index_stride, uint64_t offset);
322 void si_init_all_descriptors(struct si_context *sctx);
323 bool si_upload_vertex_buffer_descriptors(struct si_context *sctx);
324 bool si_upload_graphics_shader_descriptors(struct si_context *sctx);
325 bool si_upload_compute_shader_descriptors(struct si_context *sctx);
326 void si_release_all_descriptors(struct si_context *sctx);
327 void si_all_descriptors_begin_new_cs(struct si_context *sctx);
328 void si_upload_const_buffer(struct si_context *sctx, struct r600_resource **rbuffer,
329 const uint8_t *ptr, unsigned size, uint32_t *const_offset);
330 void si_update_all_texture_descriptors(struct si_context *sctx);
331 void si_shader_change_notify(struct si_context *sctx);
332 void si_update_needs_color_decompress_masks(struct si_context *sctx);
333 void si_emit_graphics_shader_userdata(struct si_context *sctx,
334 struct r600_atom *atom);
335 void si_emit_compute_shader_userdata(struct si_context *sctx);
336 void si_set_rw_buffer(struct si_context *sctx,
337 uint slot, const struct pipe_constant_buffer *input);
338 void si_set_active_descriptors(struct si_context *sctx, unsigned desc_idx,
339 uint64_t new_active_mask);
340 void si_set_active_descriptors_for_shader(struct si_context *sctx,
341 struct si_shader_selector *sel);
342
343 /* si_state.c */
344 struct si_shader_selector;
345
346 void si_init_atom(struct si_context *sctx, struct r600_atom *atom,
347 struct r600_atom **list_elem,
348 void (*emit_func)(struct si_context *ctx, struct r600_atom *state));
349 void si_init_state_functions(struct si_context *sctx);
350 void si_init_screen_state_functions(struct si_screen *sscreen);
351 void
352 si_make_buffer_descriptor(struct si_screen *screen, struct r600_resource *buf,
353 enum pipe_format format,
354 unsigned offset, unsigned size,
355 uint32_t *state);
356 void
357 si_make_texture_descriptor(struct si_screen *screen,
358 struct r600_texture *tex,
359 bool sampler,
360 enum pipe_texture_target target,
361 enum pipe_format pipe_format,
362 const unsigned char state_swizzle[4],
363 unsigned first_level, unsigned last_level,
364 unsigned first_layer, unsigned last_layer,
365 unsigned width, unsigned height, unsigned depth,
366 uint32_t *state,
367 uint32_t *fmask_state);
368 struct pipe_sampler_view *
369 si_create_sampler_view_custom(struct pipe_context *ctx,
370 struct pipe_resource *texture,
371 const struct pipe_sampler_view *state,
372 unsigned width0, unsigned height0,
373 unsigned force_level);
374
375 /* si_state_shader.c */
376 bool si_update_shaders(struct si_context *sctx);
377 void si_init_shader_functions(struct si_context *sctx);
378 bool si_init_shader_cache(struct si_screen *sscreen);
379 void si_destroy_shader_cache(struct si_screen *sscreen);
380 void si_init_shader_selector_async(void *job, int thread_index);
381 void si_get_active_slot_masks(const struct tgsi_shader_info *info,
382 uint32_t *const_and_shader_buffers,
383 uint64_t *samplers_and_images);
384
385 /* si_state_draw.c */
386 void si_init_ia_multi_vgt_param_table(struct si_context *sctx);
387 void si_emit_cache_flush(struct si_context *sctx);
388 void si_ce_pre_draw_synchronization(struct si_context *sctx);
389 void si_ce_post_draw_synchronization(struct si_context *sctx);
390 void si_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *dinfo);
391 void si_trace_emit(struct si_context *sctx);
392
393
394 static inline unsigned
395 si_tile_mode_index(struct r600_texture *rtex, unsigned level, bool stencil)
396 {
397 if (stencil)
398 return rtex->surface.u.legacy.stencil_tiling_index[level];
399 else
400 return rtex->surface.u.legacy.tiling_index[level];
401 }
402
403 static inline unsigned si_get_constbuf_slot(unsigned slot)
404 {
405 /* Constant buffers are in slots [16..31], ascending */
406 return SI_NUM_SHADER_BUFFERS + slot;
407 }
408
409 static inline unsigned si_get_shaderbuf_slot(unsigned slot)
410 {
411 /* shader buffers are in slots [15..0], descending */
412 return SI_NUM_SHADER_BUFFERS - 1 - slot;
413 }
414
415 static inline unsigned si_get_sampler_slot(unsigned slot)
416 {
417 /* samplers are in slots [8..39], ascending */
418 return SI_NUM_IMAGES / 2 + slot;
419 }
420
421 static inline unsigned si_get_image_slot(unsigned slot)
422 {
423 /* images are in slots [15..0] (sampler slots [7..0]), descending */
424 return SI_NUM_IMAGES - 1 - slot;
425 }
426
427 #endif