2 * Copyright 2012 Advanced Micro Devices, Inc.
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:
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
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.
24 * Christian König <christian.koenig@amd.com>
31 #include "radeon/r600_pipe_common.h"
33 #include "pipebuffer/pb_slab.h"
35 #define SI_NUM_GRAPHICS_SHADERS (PIPE_SHADER_TESS_EVAL+1)
36 #define SI_NUM_SHADERS (PIPE_SHADER_COMPUTE+1)
38 #define SI_MAX_ATTRIBS 16
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_SHADER_BUFFERS 16
47 struct si_shader_selector
;
49 struct si_state_blend
{
50 struct si_pm4_state pm4
;
51 uint32_t cb_target_mask
;
52 bool alpha_to_coverage
;
55 /* Set 0xf or 0x0 (4 bits) per render target if the following is
56 * true. ANDed with spi_shader_col_format.
58 unsigned blend_enable_4bit
;
59 unsigned need_src_alpha_4bit
;
62 struct si_state_rasterizer
{
63 struct si_pm4_state pm4
;
64 /* poly offset states for 16-bit, 24-bit, and 32-bit zbuffers */
65 struct si_pm4_state
*pm4_poly_offset
;
66 unsigned pa_sc_line_stipple
;
67 unsigned pa_cl_clip_cntl
;
68 unsigned sprite_coord_enable
:8;
69 unsigned clip_plane_enable
:8;
72 unsigned multisample_enable
:1;
73 unsigned force_persample_interp
:1;
74 unsigned line_stipple_enable
:1;
75 unsigned poly_stipple_enable
:1;
76 unsigned line_smooth
:1;
77 unsigned poly_smooth
:1;
78 unsigned uses_poly_offset
:1;
79 unsigned clamp_fragment_color
:1;
80 unsigned clamp_vertex_color
:1;
81 unsigned rasterizer_discard
:1;
82 unsigned scissor_enable
:1;
83 unsigned clip_halfz
:1;
86 struct si_dsa_stencil_ref_part
{
92 struct si_pm4_state pm4
;
94 struct si_dsa_stencil_ref_part stencil_ref
;
97 struct si_stencil_ref
{
98 struct r600_atom atom
;
99 struct pipe_stencil_ref state
;
100 struct si_dsa_stencil_ref_part dsa_part
;
103 struct si_vertex_elements
105 uint32_t instance_divisors
[SI_MAX_ATTRIBS
];
106 uint32_t rsrc_word3
[SI_MAX_ATTRIBS
];
107 uint16_t src_offset
[SI_MAX_ATTRIBS
];
108 uint8_t fix_fetch
[SI_MAX_ATTRIBS
];
109 uint8_t format_size
[SI_MAX_ATTRIBS
];
110 uint8_t vertex_buffer_index
[SI_MAX_ATTRIBS
];
113 bool uses_instance_divisors
;
115 uint16_t first_vb_use_mask
;
116 /* Vertex buffer descriptor list size aligned for optimal prefetch. */
117 uint16_t desc_list_byte_size
;
118 uint16_t instance_divisor_is_one
; /* bitmask of inputs */
119 uint16_t instance_divisor_is_fetched
; /* bitmask of inputs */
124 struct si_state_blend
*blend
;
125 struct si_state_rasterizer
*rasterizer
;
126 struct si_state_dsa
*dsa
;
127 struct si_pm4_state
*poly_offset
;
128 struct si_pm4_state
*ls
;
129 struct si_pm4_state
*hs
;
130 struct si_pm4_state
*es
;
131 struct si_pm4_state
*gs
;
132 struct si_pm4_state
*vgt_shader_config
;
133 struct si_pm4_state
*vs
;
134 struct si_pm4_state
*ps
;
136 struct si_pm4_state
*array
[0];
139 #define SI_NUM_STATES (sizeof(union si_state) / sizeof(struct si_pm4_state *))
141 union si_state_atoms
{
143 /* The order matters. */
144 struct r600_atom
*prefetch_L2
;
145 struct r600_atom
*render_cond
;
146 struct r600_atom
*streamout_begin
;
147 struct r600_atom
*streamout_enable
; /* must be after streamout_begin */
148 struct r600_atom
*framebuffer
;
149 struct r600_atom
*msaa_sample_locs
;
150 struct r600_atom
*db_render_state
;
151 struct r600_atom
*msaa_config
;
152 struct r600_atom
*sample_mask
;
153 struct r600_atom
*cb_render_state
;
154 struct r600_atom
*blend_color
;
155 struct r600_atom
*clip_regs
;
156 struct r600_atom
*clip_state
;
157 struct r600_atom
*shader_userdata
;
158 struct r600_atom
*scissors
;
159 struct r600_atom
*viewports
;
160 struct r600_atom
*stencil_ref
;
161 struct r600_atom
*spi_map
;
162 struct r600_atom
*scratch_state
;
164 struct r600_atom
*array
[0];
167 #define SI_NUM_ATOMS (sizeof(union si_state_atoms)/sizeof(struct r600_atom*))
169 struct si_shader_data
{
170 struct r600_atom atom
;
171 uint32_t sh_base
[SI_NUM_SHADERS
];
174 /* Private read-write buffer slots. */
181 SI_VS_STREAMOUT_BUF0
,
182 SI_VS_STREAMOUT_BUF1
,
183 SI_VS_STREAMOUT_BUF2
,
184 SI_VS_STREAMOUT_BUF3
,
186 SI_HS_CONST_DEFAULT_TESS_LEVELS
,
187 SI_VS_CONST_INSTANCE_DIVISORS
,
188 SI_VS_CONST_CLIP_PLANES
,
189 SI_PS_CONST_POLY_STIPPLE
,
190 SI_PS_CONST_SAMPLE_POSITIONS
,
195 /* Indices into sctx->descriptors, laid out so that gfx and compute pipelines
199 * 1 - vertex const buffers
200 * 2 - vertex shader buffers
202 * 5 - fragment const buffers
204 * 21 - compute const buffers
208 SI_SHADER_DESCS_CONST_AND_SHADER_BUFFERS
,
209 SI_SHADER_DESCS_SAMPLERS_AND_IMAGES
,
213 #define SI_DESCS_RW_BUFFERS 0
214 #define SI_DESCS_FIRST_SHADER 1
215 #define SI_DESCS_FIRST_COMPUTE (SI_DESCS_FIRST_SHADER + \
216 PIPE_SHADER_COMPUTE * SI_NUM_SHADER_DESCS)
217 #define SI_NUM_DESCS (SI_DESCS_FIRST_SHADER + \
218 SI_NUM_SHADERS * SI_NUM_SHADER_DESCS)
220 /* This represents descriptors in memory, such as buffer resources,
221 * image resources, and sampler states.
223 struct si_descriptors
{
224 /* The list of descriptors in malloc'd memory. */
226 /* The list in mapped GPU memory. */
228 /* Slots that have been changed and need to be uploaded. */
231 /* The buffer where the descriptors have been uploaded. */
232 struct r600_resource
*buffer
;
233 int buffer_offset
; /* can be negative if not using lower slots */
235 /* The size of one descriptor. */
236 ubyte element_dw_size
;
237 /* The maximum number of descriptors. */
240 /* Offset in CE RAM */
243 /* Slots allocated in CE RAM. If we get active slots outside of this
244 * range, direct uploads to memory will be used instead. This basically
245 * governs switching between onchip (CE) and offchip (upload) modes.
250 /* Slots that are used by currently-bound shaders.
251 * With CE: It determines which slots are dumped to L2.
252 * It doesn't skip uploads to CE RAM.
253 * Without CE: It determines which slots are uploaded.
255 ubyte first_active_slot
;
256 ubyte num_active_slots
;
258 /* Whether CE is used to upload this descriptor array. */
261 /* The SGPR index where the 64-bit pointer to the descriptor array will
263 ubyte shader_userdata_offset
;
266 struct si_sampler_views
{
267 struct pipe_sampler_view
*views
[SI_NUM_SAMPLERS
];
268 struct si_sampler_state
*sampler_states
[SI_NUM_SAMPLERS
];
270 /* The i-th bit is set if that element is enabled (non-NULL resource). */
271 unsigned enabled_mask
;
274 struct si_buffer_resources
{
275 struct pipe_resource
**buffers
; /* this has num_buffers elements */
277 enum radeon_bo_usage shader_usage
:4; /* READ, WRITE, or READWRITE */
278 enum radeon_bo_usage shader_usage_constbuf
:4;
279 enum radeon_bo_priority priority
:6;
280 enum radeon_bo_priority priority_constbuf
:6;
282 /* The i-th bit is set if that element is enabled (non-NULL resource). */
283 unsigned enabled_mask
;
286 #define si_pm4_block_idx(member) \
287 (offsetof(union si_state, named.member) / sizeof(struct si_pm4_state *))
289 #define si_pm4_state_changed(sctx, member) \
290 ((sctx)->queued.named.member != (sctx)->emitted.named.member)
292 #define si_pm4_bind_state(sctx, member, value) \
294 (sctx)->queued.named.member = (value); \
295 (sctx)->dirty_states |= 1 << si_pm4_block_idx(member); \
298 #define si_pm4_delete_state(sctx, member, value) \
300 if ((sctx)->queued.named.member == (value)) { \
301 (sctx)->queued.named.member = NULL; \
303 si_pm4_free_state(sctx, (struct si_pm4_state *)(value), \
304 si_pm4_block_idx(member)); \
307 /* si_descriptors.c */
308 void si_ce_save_all_descriptors_at_ib_end(struct si_context
* sctx
);
309 void si_ce_restore_all_descriptors_at_ib_start(struct si_context
*sctx
);
310 void si_ce_enable_loads(struct radeon_winsys_cs
*ib
);
311 void si_set_mutable_tex_desc_fields(struct si_screen
*sscreen
,
312 struct r600_texture
*tex
,
313 const struct legacy_surf_level
*base_level_info
,
314 unsigned base_level
, unsigned first_level
,
315 unsigned block_width
, bool is_stencil
,
317 void si_get_pipe_constant_buffer(struct si_context
*sctx
, uint shader
,
318 uint slot
, struct pipe_constant_buffer
*cbuf
);
319 void si_get_shader_buffers(struct si_context
*sctx
,
320 enum pipe_shader_type shader
,
321 uint start_slot
, uint count
,
322 struct pipe_shader_buffer
*sbuf
);
323 void si_set_ring_buffer(struct pipe_context
*ctx
, uint slot
,
324 struct pipe_resource
*buffer
,
325 unsigned stride
, unsigned num_records
,
326 bool add_tid
, bool swizzle
,
327 unsigned element_size
, unsigned index_stride
, uint64_t offset
);
328 void si_init_all_descriptors(struct si_context
*sctx
);
329 bool si_upload_vertex_buffer_descriptors(struct si_context
*sctx
);
330 bool si_upload_graphics_shader_descriptors(struct si_context
*sctx
);
331 bool si_upload_compute_shader_descriptors(struct si_context
*sctx
);
332 void si_release_all_descriptors(struct si_context
*sctx
);
333 void si_all_descriptors_begin_new_cs(struct si_context
*sctx
);
334 void si_all_resident_buffers_begin_new_cs(struct si_context
*sctx
);
335 void si_upload_const_buffer(struct si_context
*sctx
, struct r600_resource
**rbuffer
,
336 const uint8_t *ptr
, unsigned size
, uint32_t *const_offset
);
337 void si_update_all_texture_descriptors(struct si_context
*sctx
);
338 void si_shader_change_notify(struct si_context
*sctx
);
339 void si_update_needs_color_decompress_masks(struct si_context
*sctx
);
340 void si_emit_graphics_shader_userdata(struct si_context
*sctx
,
341 struct r600_atom
*atom
);
342 void si_emit_compute_shader_userdata(struct si_context
*sctx
);
343 void si_set_rw_buffer(struct si_context
*sctx
,
344 uint slot
, const struct pipe_constant_buffer
*input
);
345 void si_set_active_descriptors(struct si_context
*sctx
, unsigned desc_idx
,
346 uint64_t new_active_mask
);
347 void si_set_active_descriptors_for_shader(struct si_context
*sctx
,
348 struct si_shader_selector
*sel
);
349 bool si_bindless_descriptor_can_reclaim_slab(void *priv
,
350 struct pb_slab_entry
*entry
);
351 struct pb_slab
*si_bindless_descriptor_slab_alloc(void *priv
, unsigned heap
,
353 unsigned group_index
);
354 void si_bindless_descriptor_slab_free(void *priv
, struct pb_slab
*pslab
);
357 struct si_shader_selector
;
359 void si_init_atom(struct si_context
*sctx
, struct r600_atom
*atom
,
360 struct r600_atom
**list_elem
,
361 void (*emit_func
)(struct si_context
*ctx
, struct r600_atom
*state
));
362 void si_init_state_functions(struct si_context
*sctx
);
363 void si_init_screen_state_functions(struct si_screen
*sscreen
);
365 si_make_buffer_descriptor(struct si_screen
*screen
, struct r600_resource
*buf
,
366 enum pipe_format format
,
367 unsigned offset
, unsigned size
,
370 si_make_texture_descriptor(struct si_screen
*screen
,
371 struct r600_texture
*tex
,
373 enum pipe_texture_target target
,
374 enum pipe_format pipe_format
,
375 const unsigned char state_swizzle
[4],
376 unsigned first_level
, unsigned last_level
,
377 unsigned first_layer
, unsigned last_layer
,
378 unsigned width
, unsigned height
, unsigned depth
,
380 uint32_t *fmask_state
);
381 struct pipe_sampler_view
*
382 si_create_sampler_view_custom(struct pipe_context
*ctx
,
383 struct pipe_resource
*texture
,
384 const struct pipe_sampler_view
*state
,
385 unsigned width0
, unsigned height0
,
386 unsigned force_level
);
387 void si_update_fb_dirtiness_after_rendering(struct si_context
*sctx
);
389 /* si_state_shader.c */
390 bool si_update_shaders(struct si_context
*sctx
);
391 void si_init_shader_functions(struct si_context
*sctx
);
392 bool si_init_shader_cache(struct si_screen
*sscreen
);
393 void si_destroy_shader_cache(struct si_screen
*sscreen
);
394 void si_init_shader_selector_async(void *job
, int thread_index
);
395 void si_get_active_slot_masks(const struct tgsi_shader_info
*info
,
396 uint32_t *const_and_shader_buffers
,
397 uint64_t *samplers_and_images
);
399 /* si_state_draw.c */
400 void si_init_ia_multi_vgt_param_table(struct si_context
*sctx
);
401 void si_emit_cache_flush(struct si_context
*sctx
);
402 void si_ce_pre_draw_synchronization(struct si_context
*sctx
);
403 void si_ce_post_draw_synchronization(struct si_context
*sctx
);
404 void si_draw_vbo(struct pipe_context
*ctx
, const struct pipe_draw_info
*dinfo
);
405 void si_trace_emit(struct si_context
*sctx
);
408 static inline unsigned
409 si_tile_mode_index(struct r600_texture
*rtex
, unsigned level
, bool stencil
)
412 return rtex
->surface
.u
.legacy
.stencil_tiling_index
[level
];
414 return rtex
->surface
.u
.legacy
.tiling_index
[level
];
417 static inline unsigned si_get_constbuf_slot(unsigned slot
)
419 /* Constant buffers are in slots [16..31], ascending */
420 return SI_NUM_SHADER_BUFFERS
+ slot
;
423 static inline unsigned si_get_shaderbuf_slot(unsigned slot
)
425 /* shader buffers are in slots [15..0], descending */
426 return SI_NUM_SHADER_BUFFERS
- 1 - slot
;
429 static inline unsigned si_get_sampler_slot(unsigned slot
)
431 /* samplers are in slots [8..39], ascending */
432 return SI_NUM_IMAGES
/ 2 + slot
;
435 static inline unsigned si_get_image_slot(unsigned slot
)
437 /* images are in slots [15..0] (sampler slots [7..0]), descending */
438 return SI_NUM_IMAGES
- 1 - slot
;