aboutsummaryrefslogtreecommitdiffstats
path: root/src/gallium/drivers/radeon/radeon_winsys.h
blob: f183fe34cbbb3e12fb55f1f73b68cb75b5c7bda3 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
/*
 * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
 * Copyright 2010 Marek Olšák <maraeo@gmail.com>
 * Copyright 2018 Advanced Micro Devices, Inc.
 * All Rights Reserved.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * on the rights to use, copy, modify, merge, publish, distribute, sub
 * license, and/or sell copies of the Software, and to permit persons to whom
 * the Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
 * USE OR OTHER DEALINGS IN THE SOFTWARE. */

#ifndef RADEON_WINSYS_H
#define RADEON_WINSYS_H

/* The public winsys interface header for the radeon driver. */

/* Whether the next IB can start immediately and not wait for draws and
 * dispatches from the current IB to finish. */
#define RADEON_FLUSH_START_NEXT_GFX_IB_NOW (1u << 31)

#define RADEON_FLUSH_ASYNC_START_NEXT_GFX_IB_NOW                                                   \
   (PIPE_FLUSH_ASYNC | RADEON_FLUSH_START_NEXT_GFX_IB_NOW)

#include "amd/common/ac_gpu_info.h"
#include "amd/common/ac_surface.h"
#include "pipebuffer/pb_buffer.h"

/* Tiling flags. */
enum radeon_bo_layout
{
   RADEON_LAYOUT_LINEAR = 0,
   RADEON_LAYOUT_TILED,
   RADEON_LAYOUT_SQUARETILED,

   RADEON_LAYOUT_UNKNOWN
};

enum radeon_bo_domain
{ /* bitfield */
  RADEON_DOMAIN_GTT = 2,
  RADEON_DOMAIN_VRAM = 4,
  RADEON_DOMAIN_VRAM_GTT = RADEON_DOMAIN_VRAM | RADEON_DOMAIN_GTT,
  RADEON_DOMAIN_GDS = 8,
  RADEON_DOMAIN_OA = 16,
};

enum radeon_bo_flag
{ /* bitfield */
  RADEON_FLAG_GTT_WC = (1 << 0),
  RADEON_FLAG_NO_CPU_ACCESS = (1 << 1),
  RADEON_FLAG_NO_SUBALLOC = (1 << 2),
  RADEON_FLAG_SPARSE = (1 << 3),
  RADEON_FLAG_NO_INTERPROCESS_SHARING = (1 << 4),
  RADEON_FLAG_READ_ONLY = (1 << 5),
  RADEON_FLAG_32BIT = (1 << 6),
  RADEON_FLAG_ENCRYPTED = (1 << 7)
};

enum radeon_dependency_flag
{
   /* Add the dependency to the parallel compute IB only. */
   RADEON_DEPENDENCY_PARALLEL_COMPUTE_ONLY = 1 << 0,

   /* Instead of waiting for a job to finish execution, the dependency will
    * be signaled when the job starts execution.
    */
   RADEON_DEPENDENCY_START_FENCE = 1 << 1,
};

enum radeon_bo_usage
{ /* bitfield */
  RADEON_USAGE_READ = 2,
  RADEON_USAGE_WRITE = 4,
  RADEON_USAGE_READWRITE = RADEON_USAGE_READ | RADEON_USAGE_WRITE,

  /* The winsys ensures that the CS submission will be scheduled after
   * previously flushed CSs referencing this BO in a conflicting way.
   */
  RADEON_USAGE_SYNCHRONIZED = 8
};

enum radeon_transfer_flags
{
   /* Indicates that the caller will unmap the buffer.
    *
    * Not unmapping buffers is an important performance optimization for
    * OpenGL (avoids kernel overhead for frequently mapped buffers).
    */
   RADEON_TRANSFER_TEMPORARY = (PIPE_TRANSFER_DRV_PRV << 0),
};

#define RADEON_SPARSE_PAGE_SIZE (64 * 1024)

enum radeon_value_id
{
   RADEON_REQUESTED_VRAM_MEMORY,
   RADEON_REQUESTED_GTT_MEMORY,
   RADEON_MAPPED_VRAM,
   RADEON_MAPPED_GTT,
   RADEON_BUFFER_WAIT_TIME_NS,
   RADEON_NUM_MAPPED_BUFFERS,
   RADEON_TIMESTAMP,
   RADEON_NUM_GFX_IBS,
   RADEON_NUM_SDMA_IBS,
   RADEON_GFX_BO_LIST_COUNTER, /* number of BOs submitted in gfx IBs */
   RADEON_GFX_IB_SIZE_COUNTER,
   RADEON_NUM_BYTES_MOVED,
   RADEON_NUM_EVICTIONS,
   RADEON_NUM_VRAM_CPU_PAGE_FAULTS,
   RADEON_VRAM_USAGE,
   RADEON_VRAM_VIS_USAGE,
   RADEON_GTT_USAGE,
   RADEON_GPU_TEMPERATURE, /* DRM 2.42.0 */
   RADEON_CURRENT_SCLK,
   RADEON_CURRENT_MCLK,
   RADEON_CS_THREAD_TIME,
};

enum radeon_bo_priority
{
   /* Each group of two has the same priority. */
   RADEON_PRIO_FENCE = 0,
   RADEON_PRIO_TRACE,

   RADEON_PRIO_SO_FILLED_SIZE = 2,
   RADEON_PRIO_QUERY,

   RADEON_PRIO_IB1 = 4, /* main IB submitted to the kernel */
   RADEON_PRIO_IB2,     /* IB executed with INDIRECT_BUFFER */

   RADEON_PRIO_DRAW_INDIRECT = 6,
   RADEON_PRIO_INDEX_BUFFER,

   RADEON_PRIO_CP_DMA = 8,
   RADEON_PRIO_BORDER_COLORS,

   RADEON_PRIO_CONST_BUFFER = 10,
   RADEON_PRIO_DESCRIPTORS,

   RADEON_PRIO_SAMPLER_BUFFER = 12,
   RADEON_PRIO_VERTEX_BUFFER,

   RADEON_PRIO_SHADER_RW_BUFFER = 14,
   RADEON_PRIO_COMPUTE_GLOBAL,

   RADEON_PRIO_SAMPLER_TEXTURE = 16,
   RADEON_PRIO_SHADER_RW_IMAGE,

   RADEON_PRIO_SAMPLER_TEXTURE_MSAA = 18,
   RADEON_PRIO_COLOR_BUFFER,

   RADEON_PRIO_DEPTH_BUFFER = 20,

   RADEON_PRIO_COLOR_BUFFER_MSAA = 22,

   RADEON_PRIO_DEPTH_BUFFER_MSAA = 24,

   RADEON_PRIO_SEPARATE_META = 26,
   RADEON_PRIO_SHADER_BINARY, /* the hw can't hide instruction cache misses */

   RADEON_PRIO_SHADER_RINGS = 28,

   RADEON_PRIO_SCRATCH_BUFFER = 30,
   /* 31 is the maximum value */
};

struct winsys_handle;
struct radeon_winsys_ctx;

struct radeon_cmdbuf_chunk {
   unsigned cdw;    /* Number of used dwords. */
   unsigned max_dw; /* Maximum number of dwords. */
   uint32_t *buf;   /* The base pointer of the chunk. */
};

struct radeon_cmdbuf {
   struct radeon_cmdbuf_chunk current;
   struct radeon_cmdbuf_chunk *prev;
   unsigned num_prev; /* Number of previous chunks. */
   unsigned max_prev; /* Space in array pointed to by prev. */
   unsigned prev_dw;  /* Total number of dwords in previous chunks. */

   /* Memory usage of the buffer list. These are always 0 for preamble IBs. */
   uint64_t used_vram;
   uint64_t used_gart;
   uint64_t gpu_address;
};

/* Tiling info for display code, DRI sharing, and other data. */
struct radeon_bo_metadata {
   /* Tiling flags describing the texture layout for display code
    * and DRI sharing.
    */
   union {
      struct {
         enum radeon_bo_layout microtile;
         enum radeon_bo_layout macrotile;
         unsigned pipe_config;
         unsigned bankw;
         unsigned bankh;
         unsigned tile_split;
         unsigned mtilea;
         unsigned num_banks;
         unsigned stride;
         bool scanout;
      } legacy;
   } u;

   enum radeon_surf_mode mode;   /* Output from buffer_get_metadata */

   /* Additional metadata associated with the buffer, in bytes.
    * The maximum size is 64 * 4. This is opaque for the winsys & kernel.
    * Supported by amdgpu only.
    */
   uint32_t size_metadata;
   uint32_t metadata[64];
};

enum radeon_feature_id
{
   RADEON_FID_R300_HYPERZ_ACCESS, /* ZMask + HiZ */
   RADEON_FID_R300_CMASK_ACCESS,
};

struct radeon_bo_list_item {
   uint64_t bo_size;
   uint64_t vm_address;
   uint32_t priority_usage; /* mask of (1 << RADEON_PRIO_*) */
};

struct radeon_winsys {
   /**
    * The screen object this winsys was created for
    */
   struct pipe_screen *screen;

   /**
    * Decrement the winsys reference count.
    *
    * \param ws  The winsys this function is called for.
    * \return    True if the winsys and screen should be destroyed.
    */
   bool (*unref)(struct radeon_winsys *ws);

   /**
    * Destroy this winsys.
    *
    * \param ws        The winsys this function is called from.
    */
   void (*destroy)(struct radeon_winsys *ws);

   /**
    * Query an info structure from winsys.
    *
    * \param ws        The winsys this function is called from.
    * \param info      Return structure
    */
   void (*query_info)(struct radeon_winsys *ws, struct radeon_info *info);

   /**
    * A hint for the winsys that it should pin its execution threads to
    * a group of cores sharing a specific L3 cache if the CPU has multiple
    * L3 caches. This is needed for good multithreading performance on
    * AMD Zen CPUs.
    */
   void (*pin_threads_to_L3_cache)(struct radeon_winsys *ws, unsigned cache);

   /**************************************************************************
    * Buffer management. Buffer attributes are mostly fixed over its lifetime.
    *
    * Remember that gallium gets to choose the interface it needs, and the
    * window systems must then implement that interface (rather than the
    * other way around...).
    *************************************************************************/

   /**
    * Create a buffer object.
    *
    * \param ws        The winsys this function is called from.
    * \param size      The size to allocate.
    * \param alignment An alignment of the buffer in memory.
    * \param use_reusable_pool Whether the cache buffer manager should be used.
    * \param domain    A bitmask of the RADEON_DOMAIN_* flags.
    * \return          The created buffer object.
    */
   struct pb_buffer *(*buffer_create)(struct radeon_winsys *ws, uint64_t size, unsigned alignment,
                                      enum radeon_bo_domain domain, enum radeon_bo_flag flags);

   /**
    * Map the entire data store of a buffer object into the client's address
    * space.
    *
    * Callers are expected to unmap buffers again if and only if the
    * RADEON_TRANSFER_TEMPORARY flag is set in \p usage.
    *
    * \param buf       A winsys buffer object to map.
    * \param cs        A command stream to flush if the buffer is referenced by it.
    * \param usage     A bitmask of the PIPE_TRANSFER_* and RADEON_TRANSFER_* flags.
    * \return          The pointer at the beginning of the buffer.
    */
   void *(*buffer_map)(struct pb_buffer *buf, struct radeon_cmdbuf *cs,
                       enum pipe_transfer_usage usage);

   /**
    * Unmap a buffer object from the client's address space.
    *
    * \param buf       A winsys buffer object to unmap.
    */
   void (*buffer_unmap)(struct pb_buffer *buf);

   /**
    * Wait for the buffer and return true if the buffer is not used
    * by the device.
    *
    * The timeout of 0 will only return the status.
    * The timeout of PIPE_TIMEOUT_INFINITE will always wait until the buffer
    * is idle.
    */
   bool (*buffer_wait)(struct pb_buffer *buf, uint64_t timeout, enum radeon_bo_usage usage);

   /**
    * Return buffer metadata.
    * (tiling info for display code, DRI sharing, and other data)
    *
    * \param buf       A winsys buffer object to get the flags from.
    * \param md        Metadata
    */
   void (*buffer_get_metadata)(struct pb_buffer *buf, struct radeon_bo_metadata *md,
                               struct radeon_surf *surf);

   /**
    * Set buffer metadata.
    * (tiling info for display code, DRI sharing, and other data)
    *
    * \param buf       A winsys buffer object to set the flags for.
    * \param md        Metadata
    */
   void (*buffer_set_metadata)(struct pb_buffer *buf, struct radeon_bo_metadata *md,
                               struct radeon_surf *surf);

   /**
    * Get a winsys buffer from a winsys handle. The internal structure
    * of the handle is platform-specific and only a winsys should access it.
    *
    * \param ws        The winsys this function is called from.
    * \param whandle   A winsys handle pointer as was received from a state
    *                  tracker.
    */
   struct pb_buffer *(*buffer_from_handle)(struct radeon_winsys *ws, struct winsys_handle *whandle,
                                           unsigned vm_alignment);

   /**
    * Get a winsys buffer from a user pointer. The resulting buffer can't
    * be exported. Both pointer and size must be page aligned.
    *
    * \param ws        The winsys this function is called from.
    * \param pointer   User pointer to turn into a buffer object.
    * \param Size      Size in bytes for the new buffer.
    */
   struct pb_buffer *(*buffer_from_ptr)(struct radeon_winsys *ws, void *pointer, uint64_t size);

   /**
    * Whether the buffer was created from a user pointer.
    *
    * \param buf       A winsys buffer object
    * \return          whether \p buf was created via buffer_from_ptr
    */
   bool (*buffer_is_user_ptr)(struct pb_buffer *buf);

   /** Whether the buffer was suballocated. */
   bool (*buffer_is_suballocated)(struct pb_buffer *buf);

   /**
    * Get a winsys handle from a winsys buffer. The internal structure
    * of the handle is platform-specific and only a winsys should access it.
    *
    * \param ws        The winsys instance for which the handle is to be valid
    * \param buf       A winsys buffer object to get the handle from.
    * \param whandle   A winsys handle pointer.
    * \return          true on success.
    */
   bool (*buffer_get_handle)(struct radeon_winsys *ws, struct pb_buffer *buf,
                             struct winsys_handle *whandle);

   /**
    * Change the commitment of a (64KB-page aligned) region of the given
    * sparse buffer.
    *
    * \warning There is no automatic synchronization with command submission.
    *
    * \note Only implemented by the amdgpu winsys.
    *
    * \return false on out of memory or other failure, true on success.
    */
   bool (*buffer_commit)(struct pb_buffer *buf, uint64_t offset, uint64_t size, bool commit);

   /**
    * Return the virtual address of a buffer.
    *
    * When virtual memory is not in use, this is the offset relative to the
    * relocation base (non-zero for sub-allocated buffers).
    *
    * \param buf       A winsys buffer object
    * \return          virtual address
    */
   uint64_t (*buffer_get_virtual_address)(struct pb_buffer *buf);

   /**
    * Return the offset of this buffer relative to the relocation base.
    * This is only non-zero for sub-allocated buffers.
    *
    * This is only supported in the radeon winsys, since amdgpu uses virtual
    * addresses in submissions even for the video engines.
    *
    * \param buf      A winsys buffer object
    * \return         the offset for relocations
    */
   unsigned (*buffer_get_reloc_offset)(struct pb_buffer *buf);

   /**
    * Query the initial placement of the buffer from the kernel driver.
    */
   enum radeon_bo_domain (*buffer_get_initial_domain)(struct pb_buffer *buf);

   /**
    * Query the flags used for creation of this buffer.
    *
    * Note that for imported buffer this may be lossy since not all flags
    * are passed 1:1.
    */
   enum radeon_bo_flag (*buffer_get_flags)(struct pb_buffer *buf);

   /**************************************************************************
    * Command submission.
    *
    * Each pipe context should create its own command stream and submit
    * commands independently of other contexts.
    *************************************************************************/

   /**
    * Create a command submission context.
    * Various command streams can be submitted to the same context.
    */
   struct radeon_winsys_ctx *(*ctx_create)(struct radeon_winsys *ws);

   /**
    * Destroy a context.
    */
   void (*ctx_destroy)(struct radeon_winsys_ctx *ctx);

   /**
    * Query a GPU reset status.
    */
   enum pipe_reset_status (*ctx_query_reset_status)(struct radeon_winsys_ctx *ctx);

   /**
    * Create a command stream.
    *
    * \param ctx       The submission context
    * \param ring_type The ring type (GFX, DMA, UVD)
    * \param flush     Flush callback function associated with the command stream.
    * \param user      User pointer that will be passed to the flush callback.
    */
   struct radeon_cmdbuf *(*cs_create)(struct radeon_winsys_ctx *ctx, enum ring_type ring_type,
                                      void (*flush)(void *ctx, unsigned flags,
                                                    struct pipe_fence_handle **fence),
                                      void *flush_ctx, bool stop_exec_on_failure);

   /**
    * Add a parallel compute IB to a gfx IB. It will share the buffer list
    * and fence dependencies with the gfx IB. The gfx flush call will submit
    * both IBs at the same time.
    *
    * The compute IB doesn't have an output fence, so the primary IB has
    * to use a wait packet for synchronization.
    *
    * The returned IB is only a stream for writing packets to the new
    * IB. Calling other winsys functions with it is not allowed, not even
    * "cs_destroy". Use the gfx IB instead.
    *
    * \param cs              Gfx IB
    */
   struct radeon_cmdbuf *(*cs_add_parallel_compute_ib)(struct radeon_cmdbuf *cs,
                                                       bool uses_gds_ordered_append);

   /**
    * Destroy a command stream.
    *
    * \param cs        A command stream to destroy.
    */
   void (*cs_destroy)(struct radeon_cmdbuf *cs);

   /**
    * Add a buffer. Each buffer used by a CS must be added using this function.
    *
    * \param cs      Command stream
    * \param buf     Buffer
    * \param usage   Whether the buffer is used for read and/or write.
    * \param domain  Bitmask of the RADEON_DOMAIN_* flags.
    * \param priority  A higher number means a greater chance of being
    *                  placed in the requested domain. 15 is the maximum.
    * \return Buffer index.
    */
   unsigned (*cs_add_buffer)(struct radeon_cmdbuf *cs, struct pb_buffer *buf,
                             enum radeon_bo_usage usage, enum radeon_bo_domain domain,
                             enum radeon_bo_priority priority);

   /**
    * Return the index of an already-added buffer.
    *
    * Not supported on amdgpu. Drivers with GPUVM should not care about
    * buffer indices.
    *
    * \param cs        Command stream
    * \param buf       Buffer
    * \return          The buffer index, or -1 if the buffer has not been added.
    */
   int (*cs_lookup_buffer)(struct radeon_cmdbuf *cs, struct pb_buffer *buf);

   /**
    * Return true if there is enough memory in VRAM and GTT for the buffers
    * added so far. If the validation fails, all buffers which have
    * been added since the last call of cs_validate will be removed and
    * the CS will be flushed (provided there are still any buffers).
    *
    * \param cs        A command stream to validate.
    */
   bool (*cs_validate)(struct radeon_cmdbuf *cs);

   /**
    * Check whether the given number of dwords is available in the IB.
    * Optionally chain a new chunk of the IB if necessary and supported.
    *
    * \param cs        A command stream.
    * \param dw        Number of CS dwords requested by the caller.
    * \param force_chaining  Chain the IB into a new buffer now to discard
    *                        the CP prefetch cache (to emulate PKT3_REWIND)
    * \return true if there is enough space
    */
   bool (*cs_check_space)(struct radeon_cmdbuf *cs, unsigned dw, bool force_chaining);

   /**
    * Return the buffer list.
    *
    * This is the buffer list as passed to the kernel, i.e. it only contains
    * the parent buffers of sub-allocated buffers.
    *
    * \param cs    Command stream
    * \param list  Returned buffer list. Set to NULL to query the count only.
    * \return      The buffer count.
    */
   unsigned (*cs_get_buffer_list)(struct radeon_cmdbuf *cs, struct radeon_bo_list_item *list);

   /**
    * Flush a command stream.
    *
    * \param cs          A command stream to flush.
    * \param flags,      PIPE_FLUSH_* flags.
    * \param fence       Pointer to a fence. If non-NULL, a fence is inserted
    *                    after the CS and is returned through this parameter.
    * \return Negative POSIX error code or 0 for success.
    *         Asynchronous submissions never return an error.
    */
   int (*cs_flush)(struct radeon_cmdbuf *cs, unsigned flags, struct pipe_fence_handle **fence);

   /**
    * Create a fence before the CS is flushed.
    * The user must flush manually to complete the initializaton of the fence.
    *
    * The fence must not be used for anything except \ref cs_add_fence_dependency
    * before the flush.
    */
   struct pipe_fence_handle *(*cs_get_next_fence)(struct radeon_cmdbuf *cs);

   /**
    * Return true if a buffer is referenced by a command stream.
    *
    * \param cs        A command stream.
    * \param buf       A winsys buffer.
    */
   bool (*cs_is_buffer_referenced)(struct radeon_cmdbuf *cs, struct pb_buffer *buf,
                                   enum radeon_bo_usage usage);

   /**
    * Request access to a feature for a command stream.
    *
    * \param cs        A command stream.
    * \param fid       Feature ID, one of RADEON_FID_*
    * \param enable    Whether to enable or disable the feature.
    */
   bool (*cs_request_feature)(struct radeon_cmdbuf *cs, enum radeon_feature_id fid, bool enable);
   /**
    * Make sure all asynchronous flush of the cs have completed
    *
    * \param cs        A command stream.
    */
   void (*cs_sync_flush)(struct radeon_cmdbuf *cs);

   /**
    * Add a fence dependency to the CS, so that the CS will wait for
    * the fence before execution.
    *
    * \param dependency_flags  Bitmask of RADEON_DEPENDENCY_*
    */
   void (*cs_add_fence_dependency)(struct radeon_cmdbuf *cs, struct pipe_fence_handle *fence,
                                   unsigned dependency_flags);

   /**
    * Signal a syncobj when the CS finishes execution.
    */
   void (*cs_add_syncobj_signal)(struct radeon_cmdbuf *cs, struct pipe_fence_handle *fence);

   /**
    * Wait for the fence and return true if the fence has been signalled.
    * The timeout of 0 will only return the status.
    * The timeout of PIPE_TIMEOUT_INFINITE will always wait until the fence
    * is signalled.
    */
   bool (*fence_wait)(struct radeon_winsys *ws, struct pipe_fence_handle *fence, uint64_t timeout);

   /**
    * Reference counting for fences.
    */
   void (*fence_reference)(struct pipe_fence_handle **dst, struct pipe_fence_handle *src);

   /**
    * Create a new fence object corresponding to the given syncobj fd.
    */
   struct pipe_fence_handle *(*fence_import_syncobj)(struct radeon_winsys *ws, int fd);

   /**
    * Create a new fence object corresponding to the given sync_file.
    */
   struct pipe_fence_handle *(*fence_import_sync_file)(struct radeon_winsys *ws, int fd);

   /**
    * Return a sync_file FD corresponding to the given fence object.
    */
   int (*fence_export_sync_file)(struct radeon_winsys *ws, struct pipe_fence_handle *fence);

   /**
    * Return a sync file FD that is already signalled.
    */
   int (*export_signalled_sync_file)(struct radeon_winsys *ws);

   /**
    * Initialize surface
    *
    * \param ws        The winsys this function is called from.
    * \param tex       Input texture description
    * \param flags     Bitmask of RADEON_SURF_* flags
    * \param bpe       Bytes per pixel, it can be different for Z buffers.
    * \param mode      Preferred tile mode. (linear, 1D, or 2D)
    * \param surf      Output structure
    */
   int (*surface_init)(struct radeon_winsys *ws, const struct pipe_resource *tex, unsigned flags,
                       unsigned bpe, enum radeon_surf_mode mode, struct radeon_surf *surf);

   uint64_t (*query_value)(struct radeon_winsys *ws, enum radeon_value_id value);

   bool (*read_registers)(struct radeon_winsys *ws, unsigned reg_offset, unsigned num_registers,
                          uint32_t *out);

   /**
    * Secure context
    */
   bool (*ws_is_secure)(struct radeon_winsys *ws);
   bool (*cs_is_secure)(struct radeon_cmdbuf *cs);
   void (*cs_set_secure)(struct radeon_cmdbuf *cs, bool secure);
};

static inline bool radeon_emitted(struct radeon_cmdbuf *cs, unsigned num_dw)
{
   return cs && (cs->prev_dw + cs->current.cdw > num_dw);
}

static inline void radeon_emit(struct radeon_cmdbuf *cs, uint32_t value)
{
   cs->current.buf[cs->current.cdw++] = value;
}

static inline void radeon_emit_array(struct radeon_cmdbuf *cs, const uint32_t *values,
                                     unsigned count)
{
   memcpy(cs->current.buf + cs->current.cdw, values, count * 4);
   cs->current.cdw += count;
}

enum radeon_heap
{
   RADEON_HEAP_VRAM_NO_CPU_ACCESS,
   RADEON_HEAP_VRAM_READ_ONLY,
   RADEON_HEAP_VRAM_READ_ONLY_32BIT,
   RADEON_HEAP_VRAM_32BIT,
   RADEON_HEAP_VRAM,
   RADEON_HEAP_GTT_WC,
   RADEON_HEAP_GTT_WC_READ_ONLY,
   RADEON_HEAP_GTT_WC_READ_ONLY_32BIT,
   RADEON_HEAP_GTT_WC_32BIT,
   RADEON_HEAP_GTT,
   RADEON_MAX_SLAB_HEAPS,
   RADEON_MAX_CACHED_HEAPS = RADEON_MAX_SLAB_HEAPS,
};

static inline enum radeon_bo_domain radeon_domain_from_heap(enum radeon_heap heap)
{
   switch (heap) {
   case RADEON_HEAP_VRAM_NO_CPU_ACCESS:
   case RADEON_HEAP_VRAM_READ_ONLY:
   case RADEON_HEAP_VRAM_READ_ONLY_32BIT:
   case RADEON_HEAP_VRAM_32BIT:
   case RADEON_HEAP_VRAM:
      return RADEON_DOMAIN_VRAM;
   case RADEON_HEAP_GTT_WC:
   case RADEON_HEAP_GTT_WC_READ_ONLY:
   case RADEON_HEAP_GTT_WC_READ_ONLY_32BIT:
   case RADEON_HEAP_GTT_WC_32BIT:
   case RADEON_HEAP_GTT:
      return RADEON_DOMAIN_GTT;
   default:
      assert(0);
      return (enum radeon_bo_domain)0;
   }
}

static inline unsigned radeon_flags_from_heap(enum radeon_heap heap)
{
   unsigned flags =
      RADEON_FLAG_NO_INTERPROCESS_SHARING | (heap != RADEON_HEAP_GTT ? RADEON_FLAG_GTT_WC : 0);

   switch (heap) {
   case RADEON_HEAP_VRAM_NO_CPU_ACCESS:
      return flags | RADEON_FLAG_NO_CPU_ACCESS;

   case RADEON_HEAP_VRAM_READ_ONLY:
   case RADEON_HEAP_GTT_WC_READ_ONLY:
      return flags | RADEON_FLAG_READ_ONLY;

   case RADEON_HEAP_VRAM_READ_ONLY_32BIT:
   case RADEON_HEAP_GTT_WC_READ_ONLY_32BIT:
      return flags | RADEON_FLAG_READ_ONLY | RADEON_FLAG_32BIT;

   case RADEON_HEAP_VRAM_32BIT:
   case RADEON_HEAP_GTT_WC_32BIT:
      return flags | RADEON_FLAG_32BIT;

   case RADEON_HEAP_VRAM:
   case RADEON_HEAP_GTT_WC:
   case RADEON_HEAP_GTT:
   default:
      return flags;
   }
}

/* Return the heap index for winsys allocators, or -1 on failure. */
static inline int radeon_get_heap_index(enum radeon_bo_domain domain, enum radeon_bo_flag flags)
{
   /* VRAM implies WC (write combining) */
   assert(!(domain & RADEON_DOMAIN_VRAM) || flags & RADEON_FLAG_GTT_WC);
   /* NO_CPU_ACCESS implies VRAM only. */
   assert(!(flags & RADEON_FLAG_NO_CPU_ACCESS) || domain == RADEON_DOMAIN_VRAM);

   /* Resources with interprocess sharing don't use any winsys allocators. */
   if (!(flags & RADEON_FLAG_NO_INTERPROCESS_SHARING))
      return -1;

   /* Unsupported flags: NO_SUBALLOC, SPARSE. */
   if (flags & ~(RADEON_FLAG_GTT_WC | RADEON_FLAG_NO_CPU_ACCESS |
                 RADEON_FLAG_NO_INTERPROCESS_SHARING | RADEON_FLAG_READ_ONLY | RADEON_FLAG_32BIT))
      return -1;

   switch (domain) {
   case RADEON_DOMAIN_VRAM:
      switch (flags & (RADEON_FLAG_NO_CPU_ACCESS | RADEON_FLAG_READ_ONLY | RADEON_FLAG_32BIT)) {
      case RADEON_FLAG_NO_CPU_ACCESS | RADEON_FLAG_READ_ONLY | RADEON_FLAG_32BIT:
      case RADEON_FLAG_NO_CPU_ACCESS | RADEON_FLAG_READ_ONLY:
         assert(!"NO_CPU_ACCESS | READ_ONLY doesn't make sense");
         return -1;
      case RADEON_FLAG_NO_CPU_ACCESS | RADEON_FLAG_32BIT:
         assert(!"NO_CPU_ACCESS with 32BIT is disallowed");
         return -1;
      case RADEON_FLAG_NO_CPU_ACCESS:
         return RADEON_HEAP_VRAM_NO_CPU_ACCESS;
      case RADEON_FLAG_READ_ONLY | RADEON_FLAG_32BIT:
         return RADEON_HEAP_VRAM_READ_ONLY_32BIT;
      case RADEON_FLAG_READ_ONLY:
         return RADEON_HEAP_VRAM_READ_ONLY;
      case RADEON_FLAG_32BIT:
         return RADEON_HEAP_VRAM_32BIT;
      case 0:
         return RADEON_HEAP_VRAM;
      }
      break;
   case RADEON_DOMAIN_GTT:
      switch (flags & (RADEON_FLAG_GTT_WC | RADEON_FLAG_READ_ONLY | RADEON_FLAG_32BIT)) {
      case RADEON_FLAG_GTT_WC | RADEON_FLAG_READ_ONLY | RADEON_FLAG_32BIT:
         return RADEON_HEAP_GTT_WC_READ_ONLY_32BIT;
      case RADEON_FLAG_GTT_WC | RADEON_FLAG_READ_ONLY:
         return RADEON_HEAP_GTT_WC_READ_ONLY;
      case RADEON_FLAG_GTT_WC | RADEON_FLAG_32BIT:
         return RADEON_HEAP_GTT_WC_32BIT;
      case RADEON_FLAG_GTT_WC:
         return RADEON_HEAP_GTT_WC;
      case RADEON_FLAG_READ_ONLY | RADEON_FLAG_32BIT:
      case RADEON_FLAG_READ_ONLY:
         assert(!"READ_ONLY without WC is disallowed");
         return -1;
      case RADEON_FLAG_32BIT:
         assert(!"32BIT without WC is disallowed");
         return -1;
      case 0:
         return RADEON_HEAP_GTT;
      }
      break;
   default:
      break;
   }
   return -1;
}

#endif