radeonsi: use a bitmask for looping over dirty PM4 states
[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 24 /* 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
46 struct si_state_blend {
47 struct si_pm4_state pm4;
48 uint32_t cb_target_mask;
49 bool alpha_to_coverage;
50 bool alpha_to_one;
51 bool dual_src_blend;
52 /* Set 0xf or 0x0 (4 bits) per render target if the following is
53 * true. ANDed with spi_shader_col_format.
54 */
55 unsigned blend_enable_4bit;
56 unsigned need_src_alpha_4bit;
57 };
58
59 struct si_state_rasterizer {
60 struct si_pm4_state pm4;
61 /* poly offset states for 16-bit, 24-bit, and 32-bit zbuffers */
62 struct si_pm4_state pm4_poly_offset[3];
63 bool flatshade;
64 bool two_side;
65 bool multisample_enable;
66 bool force_persample_interp;
67 bool line_stipple_enable;
68 unsigned sprite_coord_enable;
69 unsigned pa_sc_line_stipple;
70 unsigned pa_cl_clip_cntl;
71 unsigned clip_plane_enable;
72 bool poly_stipple_enable;
73 bool line_smooth;
74 bool poly_smooth;
75 bool uses_poly_offset;
76 bool clamp_fragment_color;
77 bool rasterizer_discard;
78 bool scissor_enable;
79 bool clip_halfz;
80 };
81
82 struct si_dsa_stencil_ref_part {
83 uint8_t valuemask[2];
84 uint8_t writemask[2];
85 };
86
87 struct si_state_dsa {
88 struct si_pm4_state pm4;
89 unsigned alpha_func;
90 struct si_dsa_stencil_ref_part stencil_ref;
91 };
92
93 struct si_stencil_ref {
94 struct r600_atom atom;
95 struct pipe_stencil_ref state;
96 struct si_dsa_stencil_ref_part dsa_part;
97 };
98
99 struct si_vertex_element
100 {
101 unsigned count;
102 unsigned first_vb_use_mask;
103
104 /* Two bits per attribute indicating the size of each vector component
105 * in bytes if the size 3-workaround must be applied.
106 */
107 uint32_t fix_size3;
108 uint64_t fix_fetch;
109
110 uint32_t rsrc_word3[SI_MAX_ATTRIBS];
111 uint32_t format_size[SI_MAX_ATTRIBS];
112 struct pipe_vertex_element elements[SI_MAX_ATTRIBS];
113 };
114
115 union si_state {
116 struct {
117 struct si_state_blend *blend;
118 struct si_state_rasterizer *rasterizer;
119 struct si_state_dsa *dsa;
120 struct si_pm4_state *poly_offset;
121 struct si_pm4_state *ls;
122 struct si_pm4_state *hs;
123 struct si_pm4_state *es;
124 struct si_pm4_state *gs;
125 struct si_pm4_state *vgt_shader_config;
126 struct si_pm4_state *vs;
127 struct si_pm4_state *ps;
128 } named;
129 struct si_pm4_state *array[0];
130 };
131
132 #define SI_NUM_STATES (sizeof(union si_state) / sizeof(struct si_pm4_state *))
133
134 union si_state_atoms {
135 struct {
136 /* The order matters. */
137 struct r600_atom *prefetch_L2;
138 struct r600_atom *render_cond;
139 struct r600_atom *streamout_begin;
140 struct r600_atom *streamout_enable; /* must be after streamout_begin */
141 struct r600_atom *framebuffer;
142 struct r600_atom *msaa_sample_locs;
143 struct r600_atom *db_render_state;
144 struct r600_atom *msaa_config;
145 struct r600_atom *sample_mask;
146 struct r600_atom *cb_render_state;
147 struct r600_atom *blend_color;
148 struct r600_atom *clip_regs;
149 struct r600_atom *clip_state;
150 struct r600_atom *shader_userdata;
151 struct r600_atom *scissors;
152 struct r600_atom *viewports;
153 struct r600_atom *stencil_ref;
154 struct r600_atom *spi_map;
155 } s;
156 struct r600_atom *array[0];
157 };
158
159 #define SI_NUM_ATOMS (sizeof(union si_state_atoms)/sizeof(struct r600_atom*))
160
161 struct si_shader_data {
162 struct r600_atom atom;
163 uint32_t sh_base[SI_NUM_SHADERS];
164 };
165
166 /* Private read-write buffer slots. */
167 enum {
168 SI_HS_RING_TESS_FACTOR,
169 SI_HS_RING_TESS_OFFCHIP,
170
171 SI_ES_RING_ESGS,
172 SI_GS_RING_ESGS,
173
174 SI_RING_GSVS,
175
176 SI_VS_STREAMOUT_BUF0,
177 SI_VS_STREAMOUT_BUF1,
178 SI_VS_STREAMOUT_BUF2,
179 SI_VS_STREAMOUT_BUF3,
180
181 SI_HS_CONST_DEFAULT_TESS_LEVELS,
182 SI_VS_CONST_CLIP_PLANES,
183 SI_PS_CONST_POLY_STIPPLE,
184 SI_PS_CONST_SAMPLE_POSITIONS,
185
186 SI_NUM_RW_BUFFERS,
187 };
188
189 /* Indices into sctx->descriptors, laid out so that gfx and compute pipelines
190 * are contiguous:
191 *
192 * 0 - rw buffers
193 * 1 - vertex const buffers
194 * 2 - vertex shader buffers
195 * ...
196 * 5 - fragment const buffers
197 * ...
198 * 21 - compute const buffers
199 * ...
200 */
201 #define SI_SHADER_DESCS_CONST_BUFFERS 0
202 #define SI_SHADER_DESCS_SHADER_BUFFERS 1
203 #define SI_SHADER_DESCS_SAMPLERS 2
204 #define SI_SHADER_DESCS_IMAGES 3
205 #define SI_NUM_SHADER_DESCS 4
206
207 #define SI_DESCS_RW_BUFFERS 0
208 #define SI_DESCS_FIRST_SHADER 1
209 #define SI_DESCS_FIRST_COMPUTE (SI_DESCS_FIRST_SHADER + \
210 PIPE_SHADER_COMPUTE * SI_NUM_SHADER_DESCS)
211 #define SI_NUM_DESCS (SI_DESCS_FIRST_SHADER + \
212 SI_NUM_SHADERS * SI_NUM_SHADER_DESCS)
213
214 /* This represents descriptors in memory, such as buffer resources,
215 * image resources, and sampler states.
216 */
217 struct si_descriptors {
218 /* The list of descriptors in malloc'd memory. */
219 uint32_t *list;
220 /* The list in mapped GPU memory. */
221 uint32_t *gpu_list;
222 /* The size of one descriptor. */
223 unsigned element_dw_size;
224 /* The maximum number of descriptors. */
225 unsigned num_elements;
226
227 /* The buffer where the descriptors have been uploaded. */
228 struct r600_resource *buffer;
229 unsigned buffer_offset;
230
231 /* Offset in CE RAM */
232 unsigned ce_offset;
233
234 /* elements of the list that are changed and need to be uploaded */
235 unsigned dirty_mask;
236
237 /* Whether the CE ram is dirty and needs to be reinitialized entirely
238 * before we can do partial updates. */
239 bool ce_ram_dirty;
240
241 /* The shader userdata offset within a shader where the 64-bit pointer to the descriptor
242 * array will be stored. */
243 unsigned shader_userdata_offset;
244 };
245
246 struct si_sampler_views {
247 struct pipe_sampler_view *views[SI_NUM_SAMPLERS];
248 struct si_sampler_state *sampler_states[SI_NUM_SAMPLERS];
249
250 /* The i-th bit is set if that element is enabled (non-NULL resource). */
251 unsigned enabled_mask;
252 };
253
254 struct si_buffer_resources {
255 enum radeon_bo_usage shader_usage; /* READ, WRITE, or READWRITE */
256 enum radeon_bo_priority priority;
257 struct pipe_resource **buffers; /* this has num_buffers elements */
258
259 /* The i-th bit is set if that element is enabled (non-NULL resource). */
260 unsigned enabled_mask;
261 };
262
263 #define si_pm4_block_idx(member) \
264 (offsetof(union si_state, named.member) / sizeof(struct si_pm4_state *))
265
266 #define si_pm4_state_changed(sctx, member) \
267 ((sctx)->queued.named.member != (sctx)->emitted.named.member)
268
269 #define si_pm4_bind_state(sctx, member, value) \
270 do { \
271 (sctx)->queued.named.member = (value); \
272 (sctx)->dirty_states |= 1 << si_pm4_block_idx(member); \
273 } while(0)
274
275 #define si_pm4_delete_state(sctx, member, value) \
276 do { \
277 if ((sctx)->queued.named.member == (value)) { \
278 (sctx)->queued.named.member = NULL; \
279 } \
280 si_pm4_free_state(sctx, (struct si_pm4_state *)(value), \
281 si_pm4_block_idx(member)); \
282 } while(0)
283
284 /* si_descriptors.c */
285 void si_ce_reinitialize_all_descriptors(struct si_context *sctx);
286 void si_ce_enable_loads(struct radeon_winsys_cs *ib);
287 void si_set_mutable_tex_desc_fields(struct r600_texture *tex,
288 const struct radeon_surf_level *base_level_info,
289 unsigned base_level, unsigned first_level,
290 unsigned block_width, bool is_stencil,
291 uint32_t *state);
292 void si_get_pipe_constant_buffer(struct si_context *sctx, uint shader,
293 uint slot, struct pipe_constant_buffer *cbuf);
294 void si_get_shader_buffers(struct si_context *sctx, uint shader,
295 uint start_slot, uint count,
296 struct pipe_shader_buffer *sbuf);
297 void si_set_ring_buffer(struct pipe_context *ctx, uint slot,
298 struct pipe_resource *buffer,
299 unsigned stride, unsigned num_records,
300 bool add_tid, bool swizzle,
301 unsigned element_size, unsigned index_stride, uint64_t offset);
302 void si_init_all_descriptors(struct si_context *sctx);
303 bool si_upload_vertex_buffer_descriptors(struct si_context *sctx);
304 bool si_upload_graphics_shader_descriptors(struct si_context *sctx);
305 bool si_upload_compute_shader_descriptors(struct si_context *sctx);
306 void si_release_all_descriptors(struct si_context *sctx);
307 void si_all_descriptors_begin_new_cs(struct si_context *sctx);
308 void si_upload_const_buffer(struct si_context *sctx, struct r600_resource **rbuffer,
309 const uint8_t *ptr, unsigned size, uint32_t *const_offset);
310 void si_update_all_texture_descriptors(struct si_context *sctx);
311 void si_shader_change_notify(struct si_context *sctx);
312 void si_update_compressed_colortex_masks(struct si_context *sctx);
313 void si_emit_graphics_shader_userdata(struct si_context *sctx,
314 struct r600_atom *atom);
315 void si_emit_compute_shader_userdata(struct si_context *sctx);
316 void si_set_rw_buffer(struct si_context *sctx,
317 uint slot, const struct pipe_constant_buffer *input);
318 /* si_state.c */
319 struct si_shader_selector;
320
321 void si_init_atom(struct si_context *sctx, struct r600_atom *atom,
322 struct r600_atom **list_elem,
323 void (*emit_func)(struct si_context *ctx, struct r600_atom *state));
324 void si_init_state_functions(struct si_context *sctx);
325 void si_init_screen_state_functions(struct si_screen *sscreen);
326 void
327 si_make_buffer_descriptor(struct si_screen *screen, struct r600_resource *buf,
328 enum pipe_format format,
329 unsigned offset, unsigned size,
330 uint32_t *state);
331 void
332 si_make_texture_descriptor(struct si_screen *screen,
333 struct r600_texture *tex,
334 bool sampler,
335 enum pipe_texture_target target,
336 enum pipe_format pipe_format,
337 const unsigned char state_swizzle[4],
338 unsigned first_level, unsigned last_level,
339 unsigned first_layer, unsigned last_layer,
340 unsigned width, unsigned height, unsigned depth,
341 uint32_t *state,
342 uint32_t *fmask_state);
343 struct pipe_sampler_view *
344 si_create_sampler_view_custom(struct pipe_context *ctx,
345 struct pipe_resource *texture,
346 const struct pipe_sampler_view *state,
347 unsigned width0, unsigned height0,
348 unsigned force_level);
349
350 /* si_state_shader.c */
351 bool si_update_shaders(struct si_context *sctx);
352 void si_init_shader_functions(struct si_context *sctx);
353 bool si_init_shader_cache(struct si_screen *sscreen);
354 void si_destroy_shader_cache(struct si_screen *sscreen);
355 void si_init_shader_selector_async(void *job, int thread_index);
356
357 /* si_state_draw.c */
358 void si_emit_cache_flush(struct si_context *sctx);
359 void si_ce_pre_draw_synchronization(struct si_context *sctx);
360 void si_ce_post_draw_synchronization(struct si_context *sctx);
361 void si_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *dinfo);
362 void si_trace_emit(struct si_context *sctx);
363
364
365 static inline unsigned
366 si_tile_mode_index(struct r600_texture *rtex, unsigned level, bool stencil)
367 {
368 if (stencil)
369 return rtex->surface.stencil_tiling_index[level];
370 else
371 return rtex->surface.tiling_index[level];
372 }
373
374 #endif