aboutsummaryrefslogtreecommitdiffstats
path: root/src/mesa/drivers/dri/intel/intel_regions.c
blob: 581e5ec6b38a2d8a743b06bd72f41111990055f1 (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
/**************************************************************************
 * 
 * Copyright 2006 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.
 * 
 **************************************************************************/

/* Provide additional functionality on top of bufmgr buffers:
 *   - 2d semantics and blit operations
 *   - refcounting of buffers for multiple images in a buffer.
 *   - refcounting of buffer mappings.
 *   - some logic for moving the buffers to the best memory pools for
 *     given operations.
 *
 * Most of this is to make it easier to implement the fixed-layout
 * mipmap tree required by intel hardware in the face of GL's
 * programming interface where each image can be specifed in random
 * order and it isn't clear what layout the tree should have until the
 * last moment.
 */

#include <sys/ioctl.h>
#include <errno.h>

#include "main/hash.h"
#include "intel_context.h"
#include "intel_regions.h"
#include "intel_blit.h"
#include "intel_buffer_objects.h"
#include "intel_bufmgr.h"
#include "intel_batchbuffer.h"

#define FILE_DEBUG_FLAG DEBUG_REGION

/* This should be set to the maximum backtrace size desired.
 * Set it to 0 to disable backtrace debugging.
 */
#define DEBUG_BACKTRACE_SIZE 0

#if DEBUG_BACKTRACE_SIZE == 0
/* Use the standard debug output */
#define _DBG(...) DBG(__VA_ARGS__)
#else
/* Use backtracing debug output */
#define _DBG(...) {debug_backtrace(); DBG(__VA_ARGS__);}

/* Backtracing debug support */
#include <execinfo.h>

static void
debug_backtrace(void)
{
   void *trace[DEBUG_BACKTRACE_SIZE];
   char **strings = NULL;
   int traceSize;
   register int i;

   traceSize = backtrace(trace, DEBUG_BACKTRACE_SIZE);
   strings = backtrace_symbols(trace, traceSize);
   if (strings == NULL) {
      DBG("no backtrace:");
      return;
   }

   /* Spit out all the strings with a colon separator.  Ignore
    * the first, since we don't really care about the call
    * to debug_backtrace() itself.  Skip until the final "/" in
    * the trace to avoid really long lines.
    */
   for (i = 1; i < traceSize; i++) {
      char *p = strings[i], *slash = strings[i];
      while (*p) {
         if (*p++ == '/') {
            slash = p;
         }
      }

      DBG("%s:", slash);
   }

   /* Free up the memory, and we're done */
   free(strings);
}

#endif



/* XXX: Thread safety?
 */
GLubyte *
intel_region_map(struct intel_context *intel, struct intel_region *region)
{
   intelFlush(&intel->ctx);

   _DBG("%s %p\n", __FUNCTION__, region);
   if (!region->map_refcount++) {
      if (region->pbo)
         intel_region_cow(intel, region);

      if (region->tiling != I915_TILING_NONE)
	 drm_intel_gem_bo_map_gtt(region->buffer);
      else
	 dri_bo_map(region->buffer, GL_TRUE);
      region->map = region->buffer->virtual;
   }

   return region->map;
}

void
intel_region_unmap(struct intel_context *intel, struct intel_region *region)
{
   _DBG("%s %p\n", __FUNCTION__, region);
   if (!--region->map_refcount) {
      if (region->tiling != I915_TILING_NONE)
	 drm_intel_gem_bo_unmap_gtt(region->buffer);
      else
	 dri_bo_unmap(region->buffer);
      region->map = NULL;
   }
}

static struct intel_region *
intel_region_alloc_internal(struct intel_context *intel,
			    GLuint cpp,
			    GLuint width, GLuint height, GLuint pitch,
			    dri_bo *buffer)
{
   struct intel_region *region;

   if (buffer == NULL) {
      _DBG("%s <-- NULL\n", __FUNCTION__);
      return NULL;
   }

   region = calloc(sizeof(*region), 1);
   region->cpp = cpp;
   region->width = width;
   region->height = height;
   region->pitch = pitch;
   region->refcount = 1;
   region->buffer = buffer;

   /* Default to no tiling */
   region->tiling = I915_TILING_NONE;

   _DBG("%s <-- %p\n", __FUNCTION__, region);
   return region;
}

struct intel_region *
intel_region_alloc(struct intel_context *intel,
		   uint32_t tiling,
                   GLuint cpp, GLuint width, GLuint height, GLuint pitch,
		   GLboolean expect_accelerated_upload)
{
   dri_bo *buffer;
   struct intel_region *region;
   unsigned long flags = 0;
   unsigned long aligned_pitch;

   if (expect_accelerated_upload)
      flags |= BO_ALLOC_FOR_RENDER;

   buffer = drm_intel_bo_alloc_tiled(intel->bufmgr, "region",
				     width, height, cpp,
				     &tiling, &aligned_pitch, flags);
   /* We've already chosen a pitch as part of miptree layout.  It had
    * better be the same.
    */
   assert(aligned_pitch == pitch * cpp);

   region = intel_region_alloc_internal(intel, cpp, width, height,
					pitch, buffer);
   region->tiling = tiling;

   return region;
}

struct intel_region *
intel_region_alloc_for_handle(struct intel_context *intel,
			      GLuint cpp,
			      GLuint width, GLuint height, GLuint pitch,
			      GLuint handle, const char *name)
{
   struct intel_region *region, *dummy;
   dri_bo *buffer;
   int ret;
   uint32_t bit_6_swizzle;

   region = _mesa_HashLookup(intel->intelScreen->named_regions, handle);
   if (region != NULL) {
      dummy = NULL;
      if (region->width != width || region->height != height ||
	  region->cpp != cpp || region->pitch != pitch) {
	 fprintf(stderr,
		 "Region for name %d already exists but is not compatible\n",
		 handle);
	 return NULL;
      }
      intel_region_reference(&dummy, region);
      return dummy;
   }

   buffer = intel_bo_gem_create_from_name(intel->bufmgr, name, handle);

   region = intel_region_alloc_internal(intel, cpp,
					width, height, pitch, buffer);
   if (region == NULL)
      return region;

   ret = dri_bo_get_tiling(region->buffer, &region->tiling,
			   &bit_6_swizzle);
   if (ret != 0) {
      fprintf(stderr, "Couldn't get tiling of buffer %d (%s): %s\n",
	      handle, name, strerror(-ret));
      intel_region_release(&region);
      return NULL;
   }

   region->name = handle;
   region->screen = intel->intelScreen;
   _mesa_HashInsert(intel->intelScreen->named_regions, handle, region);

   return region;
}

void
intel_region_reference(struct intel_region **dst, struct intel_region *src)
{
   if (src)
      _DBG("%s %p %d\n", __FUNCTION__, src, src->refcount);

   assert(*dst == NULL);
   if (src) {
      src->refcount++;
      *dst = src;
   }
}

void
intel_region_release(struct intel_region **region_handle)
{
   struct intel_region *region = *region_handle;

   if (region == NULL) {
      _DBG("%s NULL\n", __FUNCTION__);
      return;
   }

   _DBG("%s %p %d\n", __FUNCTION__, region, region->refcount - 1);

   ASSERT(region->refcount > 0);
   region->refcount--;

   if (region->refcount == 0) {
      assert(region->map_refcount == 0);

      if (region->pbo)
	 region->pbo->region = NULL;
      region->pbo = NULL;
      dri_bo_unreference(region->buffer);

      if (region->name > 0)
	 _mesa_HashRemove(region->screen->named_regions, region->name);

      free(region);
   }
   *region_handle = NULL;
}

/*
 * XXX Move this into core Mesa?
 */
void
_mesa_copy_rect(GLubyte * dst,
                GLuint cpp,
                GLuint dst_pitch,
                GLuint dst_x,
                GLuint dst_y,
                GLuint width,
                GLuint height,
                const GLubyte * src,
                GLuint src_pitch, GLuint src_x, GLuint src_y)
{
   GLuint i;

   dst_pitch *= cpp;
   src_pitch *= cpp;
   dst += dst_x * cpp;
   src += src_x * cpp;
   dst += dst_y * dst_pitch;
   src += src_y * dst_pitch;
   width *= cpp;

   if (width == dst_pitch && width == src_pitch)
      memcpy(dst, src, height * width);
   else {
      for (i = 0; i < height; i++) {
         memcpy(dst, src, width);
         dst += dst_pitch;
         src += src_pitch;
      }
   }
}


/* Upload data to a rectangular sub-region.  Lots of choices how to do this:
 *
 * - memcpy by span to current destination
 * - upload data as new buffer and blit
 *
 * Currently always memcpy.
 */
void
intel_region_data(struct intel_context *intel,
                  struct intel_region *dst,
                  GLuint dst_offset,
                  GLuint dstx, GLuint dsty,
                  const void *src, GLuint src_pitch,
                  GLuint srcx, GLuint srcy, GLuint width, GLuint height)
{
   _DBG("%s\n", __FUNCTION__);

   if (intel == NULL)
      return;

   if (dst->pbo) {
      if (dstx == 0 &&
          dsty == 0 && width == dst->pitch && height == dst->height)
         intel_region_release_pbo(intel, dst);
      else
         intel_region_cow(intel, dst);
   }

   intel_prepare_render(intel);

   _mesa_copy_rect(intel_region_map(intel, dst) + dst_offset,
                   dst->cpp,
                   dst->pitch,
                   dstx, dsty, width, height, src, src_pitch, srcx, srcy);

   intel_region_unmap(intel, dst);
}

/* Copy rectangular sub-regions. Need better logic about when to
 * push buffers into AGP - will currently do so whenever possible.
 */
GLboolean
intel_region_copy(struct intel_context *intel,
                  struct intel_region *dst,
                  GLuint dst_offset,
                  GLuint dstx, GLuint dsty,
                  struct intel_region *src,
                  GLuint src_offset,
                  GLuint srcx, GLuint srcy, GLuint width, GLuint height,
		  GLboolean flip,
		  GLenum logicop)
{
   uint32_t src_pitch = src->pitch;

   _DBG("%s\n", __FUNCTION__);

   if (intel == NULL)
      return GL_FALSE;

   if (dst->pbo) {
      if (dstx == 0 &&
          dsty == 0 && width == dst->pitch && height == dst->height)
         intel_region_release_pbo(intel, dst);
      else
         intel_region_cow(intel, dst);
   }

   assert(src->cpp == dst->cpp);

   if (flip)
      src_pitch = -src_pitch;

   return intelEmitCopyBlit(intel,
			    dst->cpp,
			    src_pitch, src->buffer, src_offset, src->tiling,
			    dst->pitch, dst->buffer, dst_offset, dst->tiling,
			    srcx, srcy, dstx, dsty, width, height,
			    logicop);
}

/* Attach to a pbo, discarding our data.  Effectively zero-copy upload
 * the pbo's data.
 */
void
intel_region_attach_pbo(struct intel_context *intel,
                        struct intel_region *region,
                        struct intel_buffer_object *pbo)
{
   dri_bo *buffer;

   if (region->pbo == pbo)
      return;

   _DBG("%s %p %p\n", __FUNCTION__, region, pbo);

   /* If there is already a pbo attached, break the cow tie now.
    * Don't call intel_region_release_pbo() as that would
    * unnecessarily allocate a new buffer we would have to immediately
    * discard.
    */
   if (region->pbo) {
      region->pbo->region = NULL;
      region->pbo = NULL;
   }

   if (region->buffer) {
      dri_bo_unreference(region->buffer);
      region->buffer = NULL;
   }

   /* make sure pbo has a buffer of its own */
   buffer = intel_bufferobj_buffer(intel, pbo, INTEL_WRITE_FULL);

   region->pbo = pbo;
   region->pbo->region = region;
   dri_bo_reference(buffer);
   region->buffer = buffer;
   region->tiling = I915_TILING_NONE;
}


/* Break the COW tie to the pbo and allocate a new buffer.
 * The pbo gets to keep the data.
 */
void
intel_region_release_pbo(struct intel_context *intel,
                         struct intel_region *region)
{
   _DBG("%s %p\n", __FUNCTION__, region);
   assert(region->buffer == region->pbo->buffer);
   region->pbo->region = NULL;
   region->pbo = NULL;
   dri_bo_unreference(region->buffer);
   region->buffer = NULL;

   region->buffer = dri_bo_alloc(intel->bufmgr, "region",
				 region->pitch * region->cpp * region->height,
				 64);
}

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

   intel_region_release_pbo(intel, region);

   assert(region->cpp * region->pitch * region->height == pbo->Base.Size);

   _DBG("%s %p (%d bytes)\n", __FUNCTION__, region, pbo->Base.Size);

   /* Now blit from the texture buffer to the new buffer: 
    */

   intel_prepare_render(intel);
   ok = intelEmitCopyBlit(intel,
                          region->cpp,
                          region->pitch, pbo->buffer, 0, region->tiling,
                          region->pitch, region->buffer, 0, region->tiling,
                          0, 0, 0, 0,
                          region->pitch, region->height,
                          GL_COPY);
   assert(ok);
}

dri_bo *
intel_region_buffer(struct intel_context *intel,
                    struct intel_region *region, GLuint flag)
{
   if (region->pbo) {
      if (flag == INTEL_WRITE_PART)
         intel_region_cow(intel, region);
      else if (flag == INTEL_WRITE_FULL)
         intel_region_release_pbo(intel, region);
   }

   return region->buffer;
}