aboutsummaryrefslogtreecommitdiff
path: root/mesalib/src/gallium/auxiliary/util/u_vbuf_mgr.c
blob: 78e9df0337a91f27d0e4236ef90067358fc63479 (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
/**************************************************************************
 *
 * Copyright 2011 Marek Olšák <maraeo@gmail.com>
 * 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 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 AUTHORS AND/OR ITS 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.
 *
 **************************************************************************/

#include "util/u_vbuf_mgr.h"

#include "util/u_format.h"
#include "util/u_inlines.h"
#include "util/u_memory.h"
#include "util/u_upload_mgr.h"
#include "translate/translate.h"
#include "translate/translate_cache.h"

/* Hardware vertex fetcher limitations can be described by this structure. */
struct u_vbuf_caps {
   /* Vertex format CAPs. */
   /* TRUE if hardware supports it. */
   unsigned format_fixed32:1;    /* PIPE_FORMAT_*32*_FIXED */
   unsigned format_float16:1;    /* PIPE_FORMAT_*16*_FLOAT */
   unsigned format_float64:1;    /* PIPE_FORMAT_*64*_FLOAT */
   unsigned format_norm32:1;     /* PIPE_FORMAT_*32*NORM */
   unsigned format_scaled32:1;   /* PIPE_FORMAT_*32*SCALED */

   /* Whether vertex fetches don't have to be dword-aligned. */
   /* TRUE if hardware supports it. */
   unsigned fetch_dword_unaligned:1;
};

struct u_vbuf_mgr_elements {
   unsigned count;
   struct pipe_vertex_element ve[PIPE_MAX_ATTRIBS];

   unsigned src_format_size[PIPE_MAX_ATTRIBS];

   /* If (velem[i].src_format != native_format[i]), the vertex buffer
    * referenced by the vertex element cannot be used for rendering and
    * its vertex data must be translated to native_format[i]. */
   enum pipe_format native_format[PIPE_MAX_ATTRIBS];
   unsigned native_format_size[PIPE_MAX_ATTRIBS];

   /* This might mean two things:
    * - src_format != native_format, as discussed above.
    * - src_offset % 4 != 0 (if the caps don't allow such an offset). */
   boolean incompatible_layout;
};

struct u_vbuf_mgr_priv {
   struct u_vbuf_mgr b;
   struct u_vbuf_caps caps;
   struct pipe_context *pipe;

   struct translate_cache *translate_cache;
   unsigned translate_vb_slot;

   struct u_vbuf_mgr_elements *ve;
   void *saved_ve, *fallback_ve;
   boolean ve_binding_lock;

   boolean any_user_vbs;
   boolean incompatible_vb_layout;
};

static void u_vbuf_mgr_init_format_caps(struct u_vbuf_mgr_priv *mgr)
{
   struct pipe_screen *screen = mgr->pipe->screen;

   mgr->caps.format_fixed32 =
      screen->is_format_supported(screen, PIPE_FORMAT_R32_FIXED, PIPE_BUFFER,
                                  0, PIPE_BIND_VERTEX_BUFFER);

   mgr->caps.format_float16 =
      screen->is_format_supported(screen, PIPE_FORMAT_R16_FLOAT, PIPE_BUFFER,
                                  0, PIPE_BIND_VERTEX_BUFFER);

   mgr->caps.format_float64 =
      screen->is_format_supported(screen, PIPE_FORMAT_R64_FLOAT, PIPE_BUFFER,
                                  0, PIPE_BIND_VERTEX_BUFFER);

   mgr->caps.format_norm32 =
      screen->is_format_supported(screen, PIPE_FORMAT_R32_UNORM, PIPE_BUFFER,
                                  0, PIPE_BIND_VERTEX_BUFFER) &&
      screen->is_format_supported(screen, PIPE_FORMAT_R32_SNORM, PIPE_BUFFER,
                                  0, PIPE_BIND_VERTEX_BUFFER);

   mgr->caps.format_scaled32 =
      screen->is_format_supported(screen, PIPE_FORMAT_R32_USCALED, PIPE_BUFFER,
                                  0, PIPE_BIND_VERTEX_BUFFER) &&
      screen->is_format_supported(screen, PIPE_FORMAT_R32_SSCALED, PIPE_BUFFER,
                                  0, PIPE_BIND_VERTEX_BUFFER);
}

struct u_vbuf_mgr *
u_vbuf_mgr_create(struct pipe_context *pipe,
                  unsigned upload_buffer_size,
                  unsigned upload_buffer_alignment,
                  unsigned upload_buffer_bind,
                  enum u_fetch_alignment fetch_alignment)
{
   struct u_vbuf_mgr_priv *mgr = CALLOC_STRUCT(u_vbuf_mgr_priv);

   mgr->pipe = pipe;
   mgr->translate_cache = translate_cache_create();

   mgr->b.uploader = u_upload_create(pipe, upload_buffer_size,
                                     upload_buffer_alignment,
                                     upload_buffer_bind);

   mgr->caps.fetch_dword_unaligned =
         fetch_alignment == U_VERTEX_FETCH_BYTE_ALIGNED;

   u_vbuf_mgr_init_format_caps(mgr);

   return &mgr->b;
}

void u_vbuf_mgr_destroy(struct u_vbuf_mgr *mgrb)
{
   struct u_vbuf_mgr_priv *mgr = (struct u_vbuf_mgr_priv*)mgrb;
   unsigned i;

   for (i = 0; i < mgr->b.nr_real_vertex_buffers; i++) {
      pipe_resource_reference(&mgr->b.vertex_buffer[i].buffer, NULL);
      pipe_resource_reference(&mgr->b.real_vertex_buffer[i], NULL);
   }

   translate_cache_destroy(mgr->translate_cache);
   u_upload_destroy(mgr->b.uploader);
   FREE(mgr);
}


static void u_vbuf_translate_begin(struct u_vbuf_mgr_priv *mgr,
                                   int min_index, int max_index,
                                   boolean *upload_flushed)
{
   struct translate_key key;
   struct translate_element *te;
   unsigned tr_elem_index[PIPE_MAX_ATTRIBS];
   struct translate *tr;
   boolean vb_translated[PIPE_MAX_ATTRIBS] = {0};
   uint8_t *vb_map[PIPE_MAX_ATTRIBS] = {0}, *out_map;
   struct pipe_transfer *vb_transfer[PIPE_MAX_ATTRIBS] = {0};
   struct pipe_resource *out_buffer = NULL;
   unsigned i, num_verts, out_offset;
   struct pipe_vertex_element new_velems[PIPE_MAX_ATTRIBS];

   memset(&key, 0, sizeof(key));
   memset(tr_elem_index, 0xff, sizeof(tr_elem_index));

   /* Initialize the translate key, i.e. the recipe how vertices should be
     * translated. */
   memset(&key, 0, sizeof key);
   for (i = 0; i < mgr->ve->count; i++) {
      struct pipe_vertex_buffer *vb =
            &mgr->b.vertex_buffer[mgr->ve->ve[i].vertex_buffer_index];
      enum pipe_format output_format = mgr->ve->native_format[i];
      unsigned output_format_size = mgr->ve->native_format_size[i];

      /* Check for support. */
      if (mgr->ve->ve[i].src_format == mgr->ve->native_format[i] &&
          (mgr->caps.fetch_dword_unaligned ||
           (vb->buffer_offset % 4 == 0 &&
            vb->stride % 4 == 0 &&
            mgr->ve->ve[i].src_offset % 4 == 0))) {
         continue;
      }

      /* Workaround for translate: output floats instead of halfs. */
      switch (output_format) {
      case PIPE_FORMAT_R16_FLOAT:
         output_format = PIPE_FORMAT_R32_FLOAT;
         output_format_size = 4;
         break;
      case PIPE_FORMAT_R16G16_FLOAT:
         output_format = PIPE_FORMAT_R32G32_FLOAT;
         output_format_size = 8;
         break;
      case PIPE_FORMAT_R16G16B16_FLOAT:
         output_format = PIPE_FORMAT_R32G32B32_FLOAT;
         output_format_size = 12;
         break;
      case PIPE_FORMAT_R16G16B16A16_FLOAT:
         output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
         output_format_size = 16;
         break;
      default:;
      }

      /* Add this vertex element. */
      te = &key.element[key.nr_elements];
      /*te->type;
        te->instance_divisor;*/
      te->input_buffer = mgr->ve->ve[i].vertex_buffer_index;
      te->input_format = mgr->ve->ve[i].src_format;
      te->input_offset = mgr->ve->ve[i].src_offset;
      te->output_format = output_format;
      te->output_offset = key.output_stride;

      key.output_stride += output_format_size;
      vb_translated[mgr->ve->ve[i].vertex_buffer_index] = TRUE;
      tr_elem_index[i] = key.nr_elements;
      key.nr_elements++;
   }

   /* Get a translate object. */
   tr = translate_cache_find(mgr->translate_cache, &key);

   /* Map buffers we want to translate. */
   for (i = 0; i < mgr->b.nr_vertex_buffers; i++) {
      if (vb_translated[i]) {
         struct pipe_vertex_buffer *vb = &mgr->b.vertex_buffer[i];

         vb_map[i] = pipe_buffer_map(mgr->pipe, vb->buffer,
                                     PIPE_TRANSFER_READ, &vb_transfer[i]);

         tr->set_buffer(tr, i,
                        vb_map[i] + vb->buffer_offset + vb->stride * min_index,
                        vb->stride, ~0);
      }
   }

   /* Create and map the output buffer. */
   num_verts = max_index + 1 - min_index;

   u_upload_alloc(mgr->b.uploader,
                  key.output_stride * min_index,
                  key.output_stride * num_verts,
                  &out_offset, &out_buffer, upload_flushed,
                  (void**)&out_map);

   out_offset -= key.output_stride * min_index;

   /* Translate. */
   tr->run(tr, 0, num_verts, 0, out_map);

   /* Unmap all buffers. */
   for (i = 0; i < mgr->b.nr_vertex_buffers; i++) {
      if (vb_translated[i]) {
         pipe_buffer_unmap(mgr->pipe, vb_transfer[i]);
      }
   }

   /* Setup the new vertex buffer in the first free slot. */
   mgr->translate_vb_slot = ~0;
   for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
      if (!mgr->b.vertex_buffer[i].buffer) {
         mgr->translate_vb_slot = i;

         if (i >= mgr->b.nr_vertex_buffers) {
            mgr->b.nr_real_vertex_buffers = i+1;
         }
         break;
      }
   }

   if (mgr->translate_vb_slot != ~0) {
      /* Setup the new vertex buffer. */
      pipe_resource_reference(
            &mgr->b.real_vertex_buffer[mgr->translate_vb_slot], out_buffer);
      mgr->b.vertex_buffer[mgr->translate_vb_slot].buffer_offset = out_offset;
      mgr->b.vertex_buffer[mgr->translate_vb_slot].stride = key.output_stride;

      /* Setup new vertex elements. */
      for (i = 0; i < mgr->ve->count; i++) {
         if (tr_elem_index[i] < key.nr_elements) {
            te = &key.element[tr_elem_index[i]];
            new_velems[i].instance_divisor = mgr->ve->ve[i].instance_divisor;
            new_velems[i].src_format = te->output_format;
            new_velems[i].src_offset = te->output_offset;
            new_velems[i].vertex_buffer_index = mgr->translate_vb_slot;
         } else {
            memcpy(&new_velems[i], &mgr->ve->ve[i],
                   sizeof(struct pipe_vertex_element));
         }
      }

      mgr->fallback_ve =
            mgr->pipe->create_vertex_elements_state(mgr->pipe, mgr->ve->count,
                                                    new_velems);

      /* Preserve saved_ve. */
      mgr->ve_binding_lock = TRUE;
      mgr->pipe->bind_vertex_elements_state(mgr->pipe, mgr->fallback_ve);
      mgr->ve_binding_lock = FALSE;
   }

   pipe_resource_reference(&out_buffer, NULL);
}

static void u_vbuf_translate_end(struct u_vbuf_mgr_priv *mgr)
{
   if (mgr->fallback_ve == NULL) {
      return;
   }

   /* Restore vertex elements. */
   /* Note that saved_ve will be overwritten in bind_vertex_elements_state. */
   mgr->pipe->bind_vertex_elements_state(mgr->pipe, mgr->saved_ve);
   mgr->pipe->delete_vertex_elements_state(mgr->pipe, mgr->fallback_ve);
   mgr->fallback_ve = NULL;

   /* Delete the now-unused VBO. */
   pipe_resource_reference(&mgr->b.real_vertex_buffer[mgr->translate_vb_slot],
                           NULL);
   mgr->b.nr_real_vertex_buffers = mgr->b.nr_vertex_buffers;
}

#define FORMAT_REPLACE(what, withwhat) \
    case PIPE_FORMAT_##what: format = PIPE_FORMAT_##withwhat; break

struct u_vbuf_mgr_elements *
u_vbuf_mgr_create_vertex_elements(struct u_vbuf_mgr *mgrb,
                                  unsigned count,
                                  const struct pipe_vertex_element *attribs,
                                  struct pipe_vertex_element *native_attribs)
{
   struct u_vbuf_mgr_priv *mgr = (struct u_vbuf_mgr_priv*)mgrb;
   unsigned i;
   struct u_vbuf_mgr_elements *ve = CALLOC_STRUCT(u_vbuf_mgr_elements);

   ve->count = count;

   if (!count) {
      return ve;
   }

   memcpy(ve->ve, attribs, sizeof(struct pipe_vertex_element) * count);
   memcpy(native_attribs, attribs, sizeof(struct pipe_vertex_element) * count);

   /* Set the best native format in case the original format is not
    * supported. */
   for (i = 0; i < count; i++) {
      enum pipe_format format = ve->ve[i].src_format;

      ve->src_format_size[i] = util_format_get_blocksize(format);

      /* Choose a native format.
       * For now we don't care about the alignment, that's going to
       * be sorted out later. */
      if (!mgr->caps.format_fixed32) {
         switch (format) {
            FORMAT_REPLACE(R32_FIXED,           R32_FLOAT);
            FORMAT_REPLACE(R32G32_FIXED,        R32G32_FLOAT);
            FORMAT_REPLACE(R32G32B32_FIXED,     R32G32B32_FLOAT);
            FORMAT_REPLACE(R32G32B32A32_FIXED,  R32G32B32A32_FLOAT);
            default:;
         }
      }
      if (!mgr->caps.format_float16) {
         switch (format) {
            FORMAT_REPLACE(R16_FLOAT,           R32_FLOAT);
            FORMAT_REPLACE(R16G16_FLOAT,        R32G32_FLOAT);
            FORMAT_REPLACE(R16G16B16_FLOAT,     R32G32B32_FLOAT);
            FORMAT_REPLACE(R16G16B16A16_FLOAT,  R32G32B32A32_FLOAT);
            default:;
         }
      }
      if (!mgr->caps.format_float64) {
         switch (format) {
            FORMAT_REPLACE(R64_FLOAT,           R32_FLOAT);
            FORMAT_REPLACE(R64G64_FLOAT,        R32G32_FLOAT);
            FORMAT_REPLACE(R64G64B64_FLOAT,     R32G32B32_FLOAT);
            FORMAT_REPLACE(R64G64B64A64_FLOAT,  R32G32B32A32_FLOAT);
            default:;
         }
      }
      if (!mgr->caps.format_norm32) {
         switch (format) {
            FORMAT_REPLACE(R32_UNORM,           R32_FLOAT);
            FORMAT_REPLACE(R32G32_UNORM,        R32G32_FLOAT);
            FORMAT_REPLACE(R32G32B32_UNORM,     R32G32B32_FLOAT);
            FORMAT_REPLACE(R32G32B32A32_UNORM,  R32G32B32A32_FLOAT);
            FORMAT_REPLACE(R32_SNORM,           R32_FLOAT);
            FORMAT_REPLACE(R32G32_SNORM,        R32G32_FLOAT);
            FORMAT_REPLACE(R32G32B32_SNORM,     R32G32B32_FLOAT);
            FORMAT_REPLACE(R32G32B32A32_SNORM,  R32G32B32A32_FLOAT);
            default:;
         }
      }
      if (!mgr->caps.format_scaled32) {
         switch (format) {
            FORMAT_REPLACE(R32_USCALED,         R32_FLOAT);
            FORMAT_REPLACE(R32G32_USCALED,      R32G32_FLOAT);
            FORMAT_REPLACE(R32G32B32_USCALED,   R32G32B32_FLOAT);
            FORMAT_REPLACE(R32G32B32A32_USCALED,R32G32B32A32_FLOAT);
            FORMAT_REPLACE(R32_SSCALED,         R32_FLOAT);
            FORMAT_REPLACE(R32G32_SSCALED,      R32G32_FLOAT);
            FORMAT_REPLACE(R32G32B32_SSCALED,   R32G32B32_FLOAT);
            FORMAT_REPLACE(R32G32B32A32_SSCALED,R32G32B32A32_FLOAT);
            default:;
         }
      }

      native_attribs[i].src_format = format;
      ve->native_format[i] = format;
      ve->native_format_size[i] =
            util_format_get_blocksize(ve->native_format[i]);

      ve->incompatible_layout =
            ve->incompatible_layout ||
            ve->ve[i].src_format != ve->native_format[i] ||
            (!mgr->caps.fetch_dword_unaligned && ve->ve[i].src_offset % 4 != 0);
   }

   /* Align the formats to the size of DWORD if needed. */
   if (!mgr->caps.fetch_dword_unaligned) {
      for (i = 0; i < count; i++) {
         ve->native_format_size[i] = align(ve->native_format_size[i], 4);
      }
   }

   return ve;
}

void u_vbuf_mgr_bind_vertex_elements(struct u_vbuf_mgr *mgrb,
                                     void *cso,
                                     struct u_vbuf_mgr_elements *ve)
{
   struct u_vbuf_mgr_priv *mgr = (struct u_vbuf_mgr_priv*)mgrb;

   if (!cso) {
      return;
   }

   if (!mgr->ve_binding_lock) {
      mgr->saved_ve = cso;
      mgr->ve = ve;
   }
}

void u_vbuf_mgr_destroy_vertex_elements(struct u_vbuf_mgr *mgr,
                                        struct u_vbuf_mgr_elements *ve)
{
   FREE(ve);
}

void u_vbuf_mgr_set_vertex_buffers(struct u_vbuf_mgr *mgrb,
                                   unsigned count,
                                   const struct pipe_vertex_buffer *bufs)
{
   struct u_vbuf_mgr_priv *mgr = (struct u_vbuf_mgr_priv*)mgrb;
   unsigned i;

   mgr->any_user_vbs = FALSE;
   mgr->incompatible_vb_layout = FALSE;

   if (!mgr->caps.fetch_dword_unaligned) {
      /* Check if the strides and offsets are aligned to the size of DWORD. */
      for (i = 0; i < count; i++) {
         if (bufs[i].buffer) {
            if (bufs[i].stride % 4 != 0 ||
                bufs[i].buffer_offset % 4 != 0) {
               mgr->incompatible_vb_layout = TRUE;
               break;
            }
         }
      }
   }

   for (i = 0; i < count; i++) {
      const struct pipe_vertex_buffer *vb = &bufs[i];

      pipe_resource_reference(&mgr->b.vertex_buffer[i].buffer, vb->buffer);
      pipe_resource_reference(&mgr->b.real_vertex_buffer[i], NULL);

      if (!vb->buffer) {
         continue;
      }

      if (u_vbuf_resource(vb->buffer)->user_ptr) {
         mgr->any_user_vbs = TRUE;
         continue;
      }

      pipe_resource_reference(&mgr->b.real_vertex_buffer[i], vb->buffer);
   }

   for (; i < mgr->b.nr_real_vertex_buffers; i++) {
      pipe_resource_reference(&mgr->b.vertex_buffer[i].buffer, NULL);
      pipe_resource_reference(&mgr->b.real_vertex_buffer[i], NULL);
   }

   memcpy(mgr->b.vertex_buffer, bufs,
          sizeof(struct pipe_vertex_buffer) * count);

   mgr->b.nr_vertex_buffers = count;
   mgr->b.nr_real_vertex_buffers = count;
}

static void u_vbuf_upload_buffers(struct u_vbuf_mgr_priv *mgr,
                                  int min_index, int max_index,
                                  unsigned instance_count,
                                  boolean *upload_flushed)
{
   unsigned i, nr = mgr->ve->count;
   unsigned count = max_index + 1 - min_index;
   boolean uploaded[PIPE_MAX_ATTRIBS] = {0};

   for (i = 0; i < nr; i++) {
      unsigned index = mgr->ve->ve[i].vertex_buffer_index;
      struct pipe_vertex_buffer *vb = &mgr->b.vertex_buffer[index];

      if (vb->buffer &&
          u_vbuf_resource(vb->buffer)->user_ptr &&
          !uploaded[index]) {
         unsigned first, size;
         boolean flushed;
         unsigned instance_div = mgr->ve->ve[i].instance_divisor;

         if (instance_div) {
            first = 0;
            size = vb->stride *
                   ((instance_count + instance_div - 1) / instance_div);
         } else if (vb->stride) {
            first = vb->stride * min_index;
            size = vb->stride * count;
         } else {
            first = 0;
            size = mgr->ve->native_format_size[i];
         }

         u_upload_data(mgr->b.uploader, first, size,
                       u_vbuf_resource(vb->buffer)->user_ptr + first,
                       &vb->buffer_offset,
                       &mgr->b.real_vertex_buffer[index],
                       &flushed);

         vb->buffer_offset -= first;

         uploaded[index] = TRUE;
         *upload_flushed = *upload_flushed || flushed;
      } else {
         assert(mgr->b.real_vertex_buffer[index]);
      }
   }
}

static void u_vbuf_mgr_compute_max_index(struct u_vbuf_mgr_priv *mgr)
{
   unsigned i, nr = mgr->ve->count;

   mgr->b.max_index = ~0;

   for (i = 0; i < nr; i++) {
      struct pipe_vertex_buffer *vb =
            &mgr->b.vertex_buffer[mgr->ve->ve[i].vertex_buffer_index];
      int unused;
      unsigned max_index;

      if (!vb->buffer ||
          !vb->stride ||
          u_vbuf_resource(vb->buffer)->user_ptr) {
         continue;
      }

      /* How many bytes is unused after the last vertex.
       * width0 may be "count*stride - unused" and we have to compensate
       * for that when dividing by stride. */
      unused = vb->stride -
               (mgr->ve->ve[i].src_offset + mgr->ve->src_format_size[i]);
      assert(unused >= 0);

      /* Compute the maximum index for this vertex element. */
      max_index =
         (vb->buffer->width0 - vb->buffer_offset + (unsigned)unused) /
         vb->stride - 1;

      mgr->b.max_index = MIN2(mgr->b.max_index, max_index);
   }
}

void u_vbuf_mgr_draw_begin(struct u_vbuf_mgr *mgrb,
                           const struct pipe_draw_info *info,
                           boolean *buffers_updated,
                           boolean *uploader_flushed)
{
   struct u_vbuf_mgr_priv *mgr = (struct u_vbuf_mgr_priv*)mgrb;
   boolean bufs_updated = FALSE, upload_flushed = FALSE;
   int min_index, max_index;

   u_vbuf_mgr_compute_max_index(mgr);

   min_index = info->min_index - info->index_bias;
   if (info->max_index == ~0) {
      max_index = mgr->b.max_index;
   } else {
      max_index = MIN2(info->max_index - info->index_bias, mgr->b.max_index);
   }

   /* Translate vertices with non-native layouts or formats. */
   if (mgr->incompatible_vb_layout || mgr->ve->incompatible_layout) {
      u_vbuf_translate_begin(mgr, min_index, max_index, &upload_flushed);

      if (mgr->fallback_ve) {
         bufs_updated = TRUE;
      }
   }

   /* Upload user buffers. */
   if (mgr->any_user_vbs) {
      u_vbuf_upload_buffers(mgr, min_index, max_index, info->instance_count,
                            &upload_flushed);
      bufs_updated = TRUE;
   }

   /* Set the return values. */
   if (buffers_updated) {
      *buffers_updated = bufs_updated;
   }
   if (uploader_flushed) {
      *uploader_flushed = upload_flushed;
   }
}

void u_vbuf_mgr_draw_end(struct u_vbuf_mgr *mgrb)
{
   struct u_vbuf_mgr_priv *mgr = (struct u_vbuf_mgr_priv*)mgrb;

   if (mgr->fallback_ve) {
      u_vbuf_translate_end(mgr);
   }
}