summaryrefslogtreecommitdiff
path: root/src/mesa/drivers/dri/intel/intel_buffer_objects.c
blob: 920da1f4e075d2979918e57b47ffb73c48e9814b (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
/**************************************************************************
 * 
 * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
 * 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 TUNGSTEN GRAPHICS 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 "main/imports.h"
#include "main/mtypes.h"
#include "main/macros.h"
#include "main/bufferobj.h"

#include "intel_context.h"
#include "intel_blit.h"
#include "intel_buffer_objects.h"
#include "intel_batchbuffer.h"
#include "intel_regions.h"

static GLboolean
intel_bufferobj_unmap(GLcontext * ctx,
                      GLenum target, struct gl_buffer_object *obj);

/** Allocates a new dri_bo to store the data for the buffer object. */
static void
intel_bufferobj_alloc_buffer(struct intel_context *intel,
			     struct intel_buffer_object *intel_obj)
{
   intel_obj->buffer = dri_bo_alloc(intel->bufmgr, "bufferobj",
				    intel_obj->Base.Size, 64);
}

/**
 * There is some duplication between mesa's bufferobjects and our
 * bufmgr buffers.  Both have an integer handle and a hashtable to
 * lookup an opaque structure.  It would be nice if the handles and
 * internal structure where somehow shared.
 */
static struct gl_buffer_object *
intel_bufferobj_alloc(GLcontext * ctx, GLuint name, GLenum target)
{
   struct intel_buffer_object *obj = CALLOC_STRUCT(intel_buffer_object);

   _mesa_initialize_buffer_object(&obj->Base, name, target);

   obj->buffer = NULL;

   return &obj->Base;
}

/* Break the COW tie to the region.  The region gets to keep the data.
 */
void
intel_bufferobj_release_region(struct intel_context *intel,
                               struct intel_buffer_object *intel_obj)
{
   assert(intel_obj->region->buffer == intel_obj->buffer);
   intel_obj->region->pbo = NULL;
   intel_obj->region = NULL;

   dri_bo_unreference(intel_obj->buffer);
   intel_obj->buffer = NULL;
}

/* Break the COW tie to the region.  Both the pbo and the region end
 * up with a copy of the data.
 */
void
intel_bufferobj_cow(struct intel_context *intel,
                    struct intel_buffer_object *intel_obj)
{
   assert(intel_obj->region);
   intel_region_cow(intel, intel_obj->region);
}


/**
 * Deallocate/free a vertex/pixel buffer object.
 * Called via glDeleteBuffersARB().
 */
static void
intel_bufferobj_free(GLcontext * ctx, struct gl_buffer_object *obj)
{
   struct intel_context *intel = intel_context(ctx);
   struct intel_buffer_object *intel_obj = intel_buffer_object(obj);

   assert(intel_obj);

   /* Buffer objects are automatically unmapped when deleting according
    * to the spec, but Mesa doesn't do UnmapBuffer for us at context destroy
    * (though it does if you call glDeleteBuffers)
    */
   if (obj->Pointer)
      intel_bufferobj_unmap(ctx, 0, obj);

   _mesa_free(intel_obj->sys_buffer);
   if (intel_obj->region) {
      intel_bufferobj_release_region(intel, intel_obj);
   }
   else if (intel_obj->buffer) {
      dri_bo_unreference(intel_obj->buffer);
   }

   _mesa_free(intel_obj);
}



/**
 * Allocate space for and store data in a buffer object.  Any data that was
 * previously stored in the buffer object is lost.  If data is NULL,
 * memory will be allocated, but no copy will occur.
 * Called via glBufferDataARB().
 */
static void
intel_bufferobj_data(GLcontext * ctx,
                     GLenum target,
                     GLsizeiptrARB size,
                     const GLvoid * data,
                     GLenum usage, struct gl_buffer_object *obj)
{
   struct intel_context *intel = intel_context(ctx);
   struct intel_buffer_object *intel_obj = intel_buffer_object(obj);

   intel_obj->Base.Size = size;
   intel_obj->Base.Usage = usage;

   assert(!obj->Pointer); /* Mesa should have unmapped it */

   if (intel_obj->region)
      intel_bufferobj_release_region(intel, intel_obj);

   if (intel_obj->buffer != NULL) {
      dri_bo_unreference(intel_obj->buffer);
      intel_obj->buffer = NULL;
   }
   _mesa_free(intel_obj->sys_buffer);
   intel_obj->sys_buffer = NULL;

   if (size != 0) {
#ifdef I915
      /* On pre-965, stick VBOs in system memory, as we're always doing swtnl
       * with their contents anyway.
       */
      if (target == GL_ARRAY_BUFFER || target == GL_ELEMENT_ARRAY_BUFFER) {
	 intel_obj->sys_buffer = _mesa_malloc(size);
	 if (intel_obj->sys_buffer != NULL) {
	    if (data != NULL)
	       memcpy(intel_obj->sys_buffer, data, size);
	    return;
	 }
      }
#endif
      intel_bufferobj_alloc_buffer(intel, intel_obj);

      if (data != NULL)
	 dri_bo_subdata(intel_obj->buffer, 0, size, data);
   }
}


/**
 * Replace data in a subrange of buffer object.  If the data range
 * specified by size + offset extends beyond the end of the buffer or
 * if data is NULL, no copy is performed.
 * Called via glBufferSubDataARB().
 */
static void
intel_bufferobj_subdata(GLcontext * ctx,
                        GLenum target,
                        GLintptrARB offset,
                        GLsizeiptrARB size,
                        const GLvoid * data, struct gl_buffer_object *obj)
{
   struct intel_context *intel = intel_context(ctx);
   struct intel_buffer_object *intel_obj = intel_buffer_object(obj);

   assert(intel_obj);

   if (intel_obj->region)
      intel_bufferobj_cow(intel, intel_obj);

   if (intel_obj->sys_buffer)
      memcpy((char *)intel_obj->sys_buffer + offset, data, size);
   else
      dri_bo_subdata(intel_obj->buffer, offset, size, data);
}


/**
 * Called via glGetBufferSubDataARB().
 */
static void
intel_bufferobj_get_subdata(GLcontext * ctx,
                            GLenum target,
                            GLintptrARB offset,
                            GLsizeiptrARB size,
                            GLvoid * data, struct gl_buffer_object *obj)
{
   struct intel_buffer_object *intel_obj = intel_buffer_object(obj);

   assert(intel_obj);
   dri_bo_get_subdata(intel_obj->buffer, offset, size, data);
}



/**
 * Called via glMapBufferARB().
 */
static void *
intel_bufferobj_map(GLcontext * ctx,
                    GLenum target,
                    GLenum access, struct gl_buffer_object *obj)
{
   struct intel_context *intel = intel_context(ctx);
   struct intel_buffer_object *intel_obj = intel_buffer_object(obj);
   GLboolean read_only = (access == GL_READ_ONLY_ARB);
   GLboolean write_only = (access == GL_WRITE_ONLY_ARB);

   assert(intel_obj);

   if (intel_obj->sys_buffer) {
      obj->Pointer = intel_obj->sys_buffer;
      return obj->Pointer;
   }

   /* Flush any existing batchbuffer that might have written to this
    * buffer.
    */
   intelFlush(ctx);

   if (intel_obj->region)
      intel_bufferobj_cow(intel, intel_obj);

   if (intel_obj->buffer == NULL) {
      obj->Pointer = NULL;
      return NULL;
   }

   if (write_only && intel->intelScreen->kernel_exec_fencing) {
      drm_intel_gem_bo_map_gtt(intel_obj->buffer);
      intel_obj->mapped_gtt = GL_TRUE;
   } else {
      drm_intel_bo_map(intel_obj->buffer, !read_only);
      intel_obj->mapped_gtt = GL_FALSE;
   }

   obj->Pointer = intel_obj->buffer->virtual;
   obj->Length = obj->Size;
   obj->Offset = 0;

   return obj->Pointer;
}

/**
 * Called via glMapBufferRange().
 *
 * The goal of this extension is to allow apps to accumulate their rendering
 * at the same time as they accumulate their buffer object.  Without it,
 * you'd end up blocking on execution of rendering every time you mapped
 * the buffer to put new data in.
 *
 * We support it in 3 ways: If unsynchronized, then don't bother
 * flushing the batchbuffer before mapping the buffer, which can save blocking
 * in many cases.  If we would still block, and they allow the whole buffer
 * to be invalidated, then just allocate a new buffer to replace the old one.
 * If not, and we'd block, and they allow the subrange of the buffer to be
 * invalidated, then we can make a new little BO, let them write into that,
 * and blit it into the real BO at unmap time.
 */
static void *
intel_bufferobj_map_range(GLcontext * ctx,
			  GLenum target, GLintptr offset, GLsizeiptr length,
			  GLbitfield access, struct gl_buffer_object *obj)
{
   struct intel_context *intel = intel_context(ctx);
   struct intel_buffer_object *intel_obj = intel_buffer_object(obj);

   assert(intel_obj);

   if (intel_obj->sys_buffer) {
      obj->Pointer = intel_obj->sys_buffer + offset;
      return obj->Pointer;
   }

   if (intel_obj->region)
      intel_bufferobj_cow(intel, intel_obj);

   /* If the mapping is synchronized with other GL operations, flush
    * the batchbuffer so that GEM knows about the buffer access for later
    * syncing.
    */
   if ((access & GL_MAP_WRITE_BIT) && !(access & GL_MAP_UNSYNCHRONIZED_BIT))
      intelFlush(ctx);

   /* _mesa_MapBufferRange (GL entrypoint) sets these, but the vbo module also
    * internally uses our functions directly.
    */
   obj->Offset = offset;
   obj->Length = length;
   obj->AccessFlags = access;

   if (intel_obj->buffer == NULL) {
      obj->Pointer = NULL;
      return NULL;
   }

   /* If the user doesn't care about existing buffer contents and mapping
    * would cause us to block, then throw out the old buffer.
    */
   if (!(access & GL_MAP_UNSYNCHRONIZED_BIT) &&
       (access & GL_MAP_INVALIDATE_BUFFER_BIT) &&
       drm_intel_bo_busy(intel_obj->buffer)) {
      drm_intel_bo_unreference(intel_obj->buffer);
      intel_obj->buffer = dri_bo_alloc(intel->bufmgr, "bufferobj",
				       intel_obj->Base.Size, 64);
   }

   /* If the user is mapping a range of an active buffer object but
    * doesn't require the current contents of that range, make a new
    * BO, and we'll copy what they put in there out at unmap or
    * FlushRange time.
    */
   if ((access & GL_MAP_INVALIDATE_RANGE_BIT) &&
       drm_intel_bo_busy(intel_obj->buffer)) {
      intel_obj->range_map_bo = drm_intel_bo_alloc(intel->bufmgr,
						   "range map",
						   length, 64);
      if (!(access & GL_MAP_READ_BIT) &&
	  intel->intelScreen->kernel_exec_fencing) {
	 drm_intel_gem_bo_map_gtt(intel_obj->range_map_bo);
	 intel_obj->mapped_gtt = GL_TRUE;
      } else {
	 drm_intel_bo_map(intel_obj->range_map_bo,
			  (access & GL_MAP_WRITE_BIT) != 0);
	 intel_obj->mapped_gtt = GL_FALSE;
      }
      obj->Pointer = intel_obj->range_map_bo->virtual;
      return obj->Pointer;
   }

   if (!(access & GL_MAP_READ_BIT) &&
       intel->intelScreen->kernel_exec_fencing) {
      drm_intel_gem_bo_map_gtt(intel_obj->buffer);
      intel_obj->mapped_gtt = GL_TRUE;
   } else {
      drm_intel_bo_map(intel_obj->buffer, (access & GL_MAP_WRITE_BIT) != 0);
      intel_obj->mapped_gtt = GL_FALSE;
   }

   obj->Pointer = intel_obj->buffer->virtual + offset;
   return obj->Pointer;
}


/**
 * Called via glUnmapBuffer().
 */
static GLboolean
intel_bufferobj_unmap(GLcontext * ctx,
                      GLenum target, struct gl_buffer_object *obj)
{
   struct intel_context *intel = intel_context(ctx);
   struct intel_buffer_object *intel_obj = intel_buffer_object(obj);

   assert(intel_obj);
   assert(obj->Pointer);
   if (intel_obj->sys_buffer != NULL) {
      /* always keep the mapping around. */
   } else if (intel_obj->range_map_bo != NULL) {
      if (intel_obj->mapped_gtt) {
	 drm_intel_gem_bo_unmap_gtt(intel_obj->range_map_bo);
      } else {
	 drm_intel_bo_unmap(intel_obj->range_map_bo);
      }

      /* We ignore the FLUSH_EXPLICIT bit and the calls associated with it.
       * It would be a small win to support that, but for now we just copy
       * the whole mapped range into place.
       */
      intel_emit_linear_blit(intel,
			     intel_obj->buffer, obj->Offset,
			     intel_obj->range_map_bo, 0,
			     obj->Length);

      /* Since we've emitted some blits to buffers that will (likely) be used
       * in rendering operations in other cache domains in this batch, emit a
       * flush.  Once again, we wish for a domain tracker in libdrm to cover
       * usage inside of a batchbuffer.
       */
      intel_batchbuffer_emit_mi_flush(intel->batch);

      drm_intel_bo_unreference(intel_obj->range_map_bo);
      intel_obj->range_map_bo = NULL;
   } else if (intel_obj->buffer != NULL) {
      if (intel_obj->mapped_gtt) {
	 drm_intel_gem_bo_unmap_gtt(intel_obj->buffer);
      } else {
	 drm_intel_bo_unmap(intel_obj->buffer);
      }
   }
   obj->Pointer = NULL;
   obj->Offset = 0;
   obj->Length = 0;

   return GL_TRUE;
}

dri_bo *
intel_bufferobj_buffer(struct intel_context *intel,
                       struct intel_buffer_object *intel_obj, GLuint flag)
{
   if (intel_obj->region) {
      if (flag == INTEL_WRITE_PART)
         intel_bufferobj_cow(intel, intel_obj);
      else if (flag == INTEL_WRITE_FULL) {
         intel_bufferobj_release_region(intel, intel_obj);
	 intel_bufferobj_alloc_buffer(intel, intel_obj);
      }
   }

   if (intel_obj->buffer == NULL) {
      void *sys_buffer = intel_obj->sys_buffer;

      /* only one of buffer and sys_buffer could be non-NULL */
      intel_bufferobj_alloc_buffer(intel, intel_obj);
      intel_obj->sys_buffer = NULL;

      intel_bufferobj_subdata(&intel->ctx,
			      GL_ARRAY_BUFFER_ARB,
			      0,
			      intel_obj->Base.Size,
			      sys_buffer,
			      &intel_obj->Base);
      _mesa_free(sys_buffer);
      intel_obj->sys_buffer = NULL;
   }

   return intel_obj->buffer;
}

static void
intel_bufferobj_copy_subdata(GLcontext *ctx,
			     struct gl_buffer_object *src,
			     struct gl_buffer_object *dst,
			     GLintptr read_offset, GLintptr write_offset,
			     GLsizeiptr size)
{
   struct intel_context *intel = intel_context(ctx);
   struct intel_buffer_object *intel_src = intel_buffer_object(src);
   struct intel_buffer_object *intel_dst = intel_buffer_object(dst);
   drm_intel_bo *src_bo, *dst_bo;

   if (size == 0)
      return;

   /* If we're in system memory, just map and memcpy. */
   if (intel_src->sys_buffer || intel_dst->sys_buffer) {
      /* The same buffer may be used, but note that regions copied may
       * not overlap.
       */
      if (src == dst) {
	 char *ptr = intel_bufferobj_map(ctx, GL_COPY_WRITE_BUFFER,
					 GL_READ_WRITE, dst);
	 memcpy(ptr + write_offset, ptr + read_offset, size);
	 intel_bufferobj_unmap(ctx, GL_COPY_WRITE_BUFFER, dst);
      } else {
	 const char *src_ptr;
	 char *dst_ptr;

	 src_ptr =  intel_bufferobj_map(ctx, GL_COPY_READ_BUFFER,
					GL_READ_ONLY, src);
	 dst_ptr =  intel_bufferobj_map(ctx, GL_COPY_WRITE_BUFFER,
					GL_WRITE_ONLY, dst);

	 memcpy(dst_ptr + write_offset, src_ptr + read_offset, size);

	 intel_bufferobj_unmap(ctx, GL_COPY_READ_BUFFER, src);
	 intel_bufferobj_unmap(ctx, GL_COPY_WRITE_BUFFER, dst);
      }
   }

   /* Otherwise, we have real BOs, so blit them. */

   dst_bo = intel_bufferobj_buffer(intel, intel_dst, INTEL_WRITE_PART);
   src_bo = intel_bufferobj_buffer(intel, intel_src, INTEL_READ);

   intel_emit_linear_blit(intel,
			  dst_bo, write_offset,
			  src_bo, read_offset, size);

   /* Since we've emitted some blits to buffers that will (likely) be used
    * in rendering operations in other cache domains in this batch, emit a
    * flush.  Once again, we wish for a domain tracker in libdrm to cover
    * usage inside of a batchbuffer.
    */
   intel_batchbuffer_emit_mi_flush(intel->batch);
}

void
intelInitBufferObjectFuncs(struct dd_function_table *functions)
{
   functions->NewBufferObject = intel_bufferobj_alloc;
   functions->DeleteBuffer = intel_bufferobj_free;
   functions->BufferData = intel_bufferobj_data;
   functions->BufferSubData = intel_bufferobj_subdata;
   functions->GetBufferSubData = intel_bufferobj_get_subdata;
   functions->MapBuffer = intel_bufferobj_map;
   functions->MapBufferRange = intel_bufferobj_map_range;
   functions->UnmapBuffer = intel_bufferobj_unmap;
   functions->CopyBufferSubData = intel_bufferobj_copy_subdata;
}