2 * Copyright 2010 Jerome Glisse <glisse@freedesktop.org>
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.
29 #include "util/u_double_list.h"
30 #include "util/u_inlines.h"
32 #define RADEON_CTX_MAX_PM4 (64 * 1024 / 4)
34 #define R600_ERR(fmt, args...) \
35 fprintf(stderr, "EE %s:%d %s - "fmt, __FILE__, __LINE__, __func__, ##args)
81 struct r600_tiling_info
{
82 unsigned num_channels
;
87 enum radeon_family
r600_get_family(struct radeon
*rw
);
88 enum chip_class
r600_get_family_class(struct radeon
*radeon
);
89 struct r600_tiling_info
*r600_get_tiling_info(struct radeon
*radeon
);
90 unsigned r600_get_clock_crystal_freq(struct radeon
*radeon
);
91 unsigned r600_get_minor_version(struct radeon
*radeon
);
92 unsigned r600_get_num_backends(struct radeon
*radeon
);
93 unsigned r600_get_num_tile_pipes(struct radeon
*radeon
);
94 unsigned r600_get_backend_map(struct radeon
*radeon
);
98 struct radeon_winsys_cs
;
100 struct r600_bo
*r600_bo(struct radeon
*radeon
,
101 unsigned size
, unsigned alignment
,
102 unsigned binding
, unsigned usage
);
103 struct r600_bo
*r600_bo_handle(struct radeon
*radeon
, struct winsys_handle
*whandle
,
104 unsigned *stride
, unsigned *array_mode
);
105 void *r600_bo_map(struct radeon
*radeon
, struct r600_bo
*bo
, struct radeon_winsys_cs
*cs
, unsigned usage
);
106 void r600_bo_unmap(struct radeon
*radeon
, struct r600_bo
*bo
);
107 boolean
r600_bo_get_winsys_handle(struct radeon
*radeon
, struct r600_bo
*pb_bo
,
108 unsigned stride
, struct winsys_handle
*whandle
);
110 void r600_bo_destroy(struct radeon
*radeon
, struct r600_bo
*bo
);
112 /* this relies on the pipe_reference being the first member of r600_bo */
113 static INLINE
void r600_bo_reference(struct radeon
*radeon
, struct r600_bo
**dst
, struct r600_bo
*src
)
115 struct r600_bo
*old
= *dst
;
117 if (pipe_reference((struct pipe_reference
*)(*dst
), (struct pipe_reference
*)src
)) {
118 r600_bo_destroy(radeon
, old
);
124 /* R600/R700 STATES */
125 #define R600_GROUP_MAX 16
126 #define R600_BLOCK_MAX_BO 32
127 #define R600_BLOCK_MAX_REG 128
129 /* each range covers 9 bits of dword space = 512 dwords = 2k bytes */
130 /* there is a block entry for each register so 512 blocks */
131 /* we have no registers to read/write below 0x8000 (0x2000 in dw space) */
132 /* we use some fake offsets at 0x40000 to do evergreen sampler borders so take 0x42000 as a max bound*/
133 #define RANGE_OFFSET_START 0x8000
135 #define NUM_RANGES (0x42000 - RANGE_OFFSET_START) / (4 << HASH_SHIFT) /* 128 << 9 = 64k */
137 #define CTX_RANGE_ID(offset) ((((offset - RANGE_OFFSET_START) >> 2) >> HASH_SHIFT) & 255)
138 #define CTX_BLOCK_ID(offset) (((offset - RANGE_OFFSET_START) >> 2) & ((1 << HASH_SHIFT) - 1))
140 struct r600_pipe_reg
{
143 struct r600_block
*block
;
148 struct r600_pipe_state
{
151 struct r600_pipe_reg regs
[R600_BLOCK_MAX_REG
];
154 struct r600_pipe_resource_state
{
157 struct r600_bo
*bo
[2];
160 #define R600_BLOCK_STATUS_ENABLED (1 << 0)
161 #define R600_BLOCK_STATUS_DIRTY (1 << 1)
162 #define R600_BLOCK_STATUS_RESOURCE_DIRTY (1 << 2)
164 #define R600_BLOCK_STATUS_RESOURCE_VERTEX (1 << 3)
166 struct r600_block_reloc
{
168 unsigned flush_flags
;
170 unsigned bo_pm4_index
;
174 struct list_head list
;
175 struct list_head enable_list
;
178 unsigned start_offset
;
179 unsigned pm4_ndwords
;
180 unsigned pm4_flush_ndwords
;
185 u32 pm4
[R600_BLOCK_MAX_REG
];
186 unsigned pm4_bo_index
[R600_BLOCK_MAX_REG
];
187 struct r600_block_reloc reloc
[R600_BLOCK_MAX_BO
];
191 struct r600_block
**blocks
;
199 /* The kind of query */
201 /* Offset of the first result for current query */
202 unsigned results_start
;
203 /* Offset of the next free result after current query data */
204 unsigned results_end
;
205 /* Size of the result */
206 unsigned result_size
;
207 /* Count of new queries started in one stream without flushing */
208 unsigned queries_emitted
;
211 /* The buffer where query results are stored. It's used as a ring,
212 * data blocks for current query are stored sequentially from
213 * results_start to results_end, with wrapping on the buffer end */
214 struct r600_bo
*buffer
;
215 unsigned buffer_size
;
216 /* linked list of queries */
217 struct list_head list
;
220 #define R600_QUERY_STATE_STARTED (1 << 0)
221 #define R600_QUERY_STATE_ENDED (1 << 1)
222 #define R600_QUERY_STATE_SUSPENDED (1 << 2)
223 #define R600_QUERY_STATE_FLUSHED (1 << 3)
225 #define R600_CONTEXT_DRAW_PENDING (1 << 0)
226 #define R600_CONTEXT_DST_CACHES_DIRTY (1 << 1)
227 #define R600_CONTEXT_CHECK_EVENT_FLUSH (1 << 2)
229 struct r600_context
{
230 struct radeon
*radeon
;
231 struct radeon_winsys_cs
*cs
;
233 struct r600_range
*range
;
235 struct r600_block
**blocks
;
236 struct list_head dirty
;
237 struct list_head resource_dirty
;
238 struct list_head enable_list
;
239 unsigned pm4_ndwords
;
240 unsigned pm4_dirty_cdwords
;
241 unsigned ctx_pm4_ndwords
;
242 unsigned init_dwords
;
245 struct radeon_bo
**bo
;
248 unsigned pm4_cdwords
;
250 struct list_head query_list
;
251 unsigned num_query_running
;
252 unsigned backend_mask
;
253 unsigned max_db
; /* for OQ */
254 unsigned num_dest_buffers
;
256 boolean predicate_drawing
;
257 struct r600_range ps_resources
;
258 struct r600_range vs_resources
;
259 struct r600_range fs_resources
;
260 int num_ps_resources
, num_vs_resources
, num_fs_resources
;
261 boolean have_depth_texture
, have_depth_fb
;
266 u32 vgt_num_instances
;
268 u32 vgt_draw_initiator
;
269 u32 indices_bo_offset
;
270 struct r600_bo
*indices
;
273 void r600_get_backend_mask(struct r600_context
*ctx
);
274 int r600_context_init(struct r600_context
*ctx
, struct radeon
*radeon
);
275 void r600_context_fini(struct r600_context
*ctx
);
276 void r600_context_pipe_state_set(struct r600_context
*ctx
, struct r600_pipe_state
*state
);
277 void r600_context_pipe_state_set_ps_resource(struct r600_context
*ctx
, struct r600_pipe_resource_state
*state
, unsigned rid
);
278 void r600_context_pipe_state_set_vs_resource(struct r600_context
*ctx
, struct r600_pipe_resource_state
*state
, unsigned rid
);
279 void r600_context_pipe_state_set_fs_resource(struct r600_context
*ctx
, struct r600_pipe_resource_state
*state
, unsigned rid
);
280 void r600_context_pipe_state_set_ps_sampler(struct r600_context
*ctx
, struct r600_pipe_state
*state
, unsigned id
);
281 void r600_context_pipe_state_set_vs_sampler(struct r600_context
*ctx
, struct r600_pipe_state
*state
, unsigned id
);
282 void r600_context_flush(struct r600_context
*ctx
, unsigned flags
);
283 void r600_context_draw(struct r600_context
*ctx
, const struct r600_draw
*draw
);
285 struct r600_query
*r600_context_query_create(struct r600_context
*ctx
, unsigned query_type
);
286 void r600_context_query_destroy(struct r600_context
*ctx
, struct r600_query
*query
);
287 boolean
r600_context_query_result(struct r600_context
*ctx
,
288 struct r600_query
*query
,
289 boolean wait
, void *vresult
);
290 void r600_query_begin(struct r600_context
*ctx
, struct r600_query
*query
);
291 void r600_query_end(struct r600_context
*ctx
, struct r600_query
*query
);
292 void r600_context_queries_suspend(struct r600_context
*ctx
);
293 void r600_context_queries_resume(struct r600_context
*ctx
, boolean flushed
);
294 void r600_query_predication(struct r600_context
*ctx
, struct r600_query
*query
, int operation
,
296 void r600_context_emit_fence(struct r600_context
*ctx
, struct r600_bo
*fence
,
297 unsigned offset
, unsigned value
);
298 void r600_context_flush_all(struct r600_context
*ctx
, unsigned flush_flags
);
299 void r600_context_flush_dest_caches(struct r600_context
*ctx
);
301 int evergreen_context_init(struct r600_context
*ctx
, struct radeon
*radeon
);
302 void evergreen_context_draw(struct r600_context
*ctx
, const struct r600_draw
*draw
);
303 void evergreen_context_flush_dest_caches(struct r600_context
*ctx
);
304 void evergreen_context_pipe_state_set_ps_resource(struct r600_context
*ctx
, struct r600_pipe_resource_state
*state
, unsigned rid
);
305 void evergreen_context_pipe_state_set_vs_resource(struct r600_context
*ctx
, struct r600_pipe_resource_state
*state
, unsigned rid
);
306 void evergreen_context_pipe_state_set_fs_resource(struct r600_context
*ctx
, struct r600_pipe_resource_state
*state
, unsigned rid
);
307 void evergreen_context_pipe_state_set_ps_sampler(struct r600_context
*ctx
, struct r600_pipe_state
*state
, unsigned id
);
308 void evergreen_context_pipe_state_set_vs_sampler(struct r600_context
*ctx
, struct r600_pipe_state
*state
, unsigned id
);
310 struct radeon
*radeon_destroy(struct radeon
*radeon
);
312 void _r600_pipe_state_add_reg(struct r600_context
*ctx
,
313 struct r600_pipe_state
*state
,
314 u32 offset
, u32 value
, u32 mask
,
315 u32 range_id
, u32 block_id
,
318 void r600_pipe_state_add_reg_noblock(struct r600_pipe_state
*state
,
319 u32 offset
, u32 value
, u32 mask
,
321 #define r600_pipe_state_add_reg(state, offset, value, mask, bo) _r600_pipe_state_add_reg(&rctx->ctx, state, offset, value, mask, CTX_RANGE_ID(offset), CTX_BLOCK_ID(offset), bo)
323 static inline void r600_pipe_state_mod_reg(struct r600_pipe_state
*state
,
326 state
->regs
[state
->nregs
].value
= value
;
330 static inline void r600_pipe_state_mod_reg_bo(struct r600_pipe_state
*state
,
331 u32 value
, struct r600_bo
*bo
)
333 state
->regs
[state
->nregs
].value
= value
;
334 state
->regs
[state
->nregs
].bo
= bo
;