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
|
/**********************************************************
* Copyright 2008-2009 VMware, 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 the rights to use, copy,
* modify, merge, publish, distribute, sublicense, 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 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
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
* 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 "svga_cmd.h"
#include "pipe/p_state.h"
#include "pipe/p_defines.h"
#include "util/u_inlines.h"
#include "os/os_thread.h"
#include "util/u_format.h"
#include "util/u_math.h"
#include "util/u_memory.h"
#include "svga_format.h"
#include "svga_screen.h"
#include "svga_context.h"
#include "svga_resource_texture.h"
#include "svga_resource_buffer.h"
#include "svga_sampler_view.h"
#include "svga_winsys.h"
#include "svga_debug.h"
/* XXX: This isn't a real hardware flag, but just a hack for kernel to
* know about primary surfaces. Find a better way to accomplish this.
*/
#define SVGA3D_SURFACE_HINT_SCANOUT (1 << 9)
static INLINE void
svga_transfer_dma_band(struct svga_context *svga,
struct svga_transfer *st,
SVGA3dTransferType transfer,
unsigned y, unsigned h, unsigned srcy,
SVGA3dSurfaceDMAFlags flags)
{
struct svga_texture *texture = svga_texture(st->base.resource);
SVGA3dCopyBox box;
enum pipe_error ret;
box.x = st->base.box.x;
box.y = y;
box.z = st->base.box.z;
box.w = st->base.box.width;
box.h = h;
box.d = 1;
box.srcx = 0;
box.srcy = srcy;
box.srcz = 0;
if (st->base.resource->target == PIPE_TEXTURE_CUBE) {
st->face = st->base.box.z;
box.z = 0;
}
else
st->face = 0;
SVGA_DBG(DEBUG_DMA, "dma %s sid %p, face %u, (%u, %u, %u) - (%u, %u, %u), %ubpp\n",
transfer == SVGA3D_WRITE_HOST_VRAM ? "to" : "from",
texture->handle,
st->face,
st->base.box.x,
y,
box.z,
st->base.box.x + st->base.box.width,
y + h,
box.z + 1,
util_format_get_blocksize(texture->b.b.format) * 8 /
(util_format_get_blockwidth(texture->b.b.format)*util_format_get_blockheight(texture->b.b.format)));
ret = SVGA3D_SurfaceDMA(svga->swc, st, transfer, &box, 1, flags);
if(ret != PIPE_OK) {
svga_context_flush(svga, NULL);
ret = SVGA3D_SurfaceDMA(svga->swc, st, transfer, &box, 1, flags);
assert(ret == PIPE_OK);
}
}
static INLINE void
svga_transfer_dma(struct svga_context *svga,
struct svga_transfer *st,
SVGA3dTransferType transfer,
SVGA3dSurfaceDMAFlags flags)
{
struct svga_texture *texture = svga_texture(st->base.resource);
struct svga_screen *screen = svga_screen(texture->b.b.screen);
struct svga_winsys_screen *sws = screen->sws;
struct pipe_fence_handle *fence = NULL;
if (transfer == SVGA3D_READ_HOST_VRAM) {
SVGA_DBG(DEBUG_PERF, "%s: readback transfer\n", __FUNCTION__);
}
/* Ensure any pending operations on host surfaces are queued on the command
* buffer first.
*/
svga_surfaces_flush( svga );
if(!st->swbuf) {
/* Do the DMA transfer in a single go */
svga_transfer_dma_band(svga, st, transfer,
st->base.box.y, st->base.box.height, 0,
flags);
if(transfer == SVGA3D_READ_HOST_VRAM) {
svga_context_flush(svga, &fence);
sws->fence_finish(sws, fence, 0);
sws->fence_reference(sws, &fence, NULL);
}
}
else {
unsigned y, h, srcy;
unsigned blockheight = util_format_get_blockheight(st->base.resource->format);
h = st->hw_nblocksy * blockheight;
srcy = 0;
for(y = 0; y < st->base.box.height; y += h) {
unsigned offset, length;
void *hw, *sw;
if (y + h > st->base.box.height)
h = st->base.box.height - y;
/* Transfer band must be aligned to pixel block boundaries */
assert(y % blockheight == 0);
assert(h % blockheight == 0);
offset = y * st->base.stride / blockheight;
length = h * st->base.stride / blockheight;
sw = (uint8_t *)st->swbuf + offset;
if (transfer == SVGA3D_WRITE_HOST_VRAM) {
unsigned usage = PIPE_TRANSFER_WRITE;
/* Wait for the previous DMAs to complete */
/* TODO: keep one DMA (at half the size) in the background */
if (y) {
svga_context_flush(svga, NULL);
usage |= PIPE_TRANSFER_DISCARD_WHOLE_RESOURCE;
}
hw = sws->buffer_map(sws, st->hwbuf, usage);
assert(hw);
if (hw) {
memcpy(hw, sw, length);
sws->buffer_unmap(sws, st->hwbuf);
}
}
svga_transfer_dma_band(svga, st, transfer, y, h, srcy, flags);
/*
* Prevent the texture contents to be discarded on the next band
* upload.
*/
flags.discard = FALSE;
if(transfer == SVGA3D_READ_HOST_VRAM) {
svga_context_flush(svga, &fence);
sws->fence_finish(sws, fence, 0);
hw = sws->buffer_map(sws, st->hwbuf, PIPE_TRANSFER_READ);
assert(hw);
if(hw) {
memcpy(sw, hw, length);
sws->buffer_unmap(sws, st->hwbuf);
}
}
}
}
}
static boolean
svga_texture_get_handle(struct pipe_screen *screen,
struct pipe_resource *texture,
struct winsys_handle *whandle)
{
struct svga_winsys_screen *sws = svga_winsys_screen(texture->screen);
unsigned stride;
assert(svga_texture(texture)->key.cachable == 0);
svga_texture(texture)->key.cachable = 0;
stride = util_format_get_nblocksx(texture->format, texture->width0) *
util_format_get_blocksize(texture->format);
return sws->surface_get_handle(sws, svga_texture(texture)->handle, stride, whandle);
}
static void
svga_texture_destroy(struct pipe_screen *screen,
struct pipe_resource *pt)
{
struct svga_screen *ss = svga_screen(screen);
struct svga_texture *tex = (struct svga_texture *)pt;
ss->texture_timestamp++;
svga_sampler_view_reference(&tex->cached_view, NULL);
/*
DBG("%s deleting %p\n", __FUNCTION__, (void *) tex);
*/
SVGA_DBG(DEBUG_DMA, "unref sid %p (texture)\n", tex->handle);
svga_screen_surface_destroy(ss, &tex->key, &tex->handle);
FREE(tex);
}
/* XXX: Still implementing this as if it was a screen function, but
* can now modify it to queue transfers on the context.
*/
static struct pipe_transfer *
svga_texture_get_transfer(struct pipe_context *pipe,
struct pipe_resource *texture,
unsigned level,
unsigned usage,
const struct pipe_box *box)
{
struct svga_context *svga = svga_context(pipe);
struct svga_screen *ss = svga_screen(pipe->screen);
struct svga_winsys_screen *sws = ss->sws;
struct svga_transfer *st;
unsigned nblocksx = util_format_get_nblocksx(texture->format, box->width);
unsigned nblocksy = util_format_get_nblocksy(texture->format, box->height);
/* We can't map texture storage directly */
if (usage & (PIPE_TRANSFER_MAP_DIRECTLY | PIPE_TRANSFER_MAP_PERMANENTLY))
return NULL;
assert(box->depth == 1);
st = CALLOC_STRUCT(svga_transfer);
if (!st)
return NULL;
pipe_resource_reference(&st->base.resource, texture);
st->base.level = level;
st->base.usage = usage;
st->base.box = *box;
st->base.stride = nblocksx*util_format_get_blocksize(texture->format);
st->base.layer_stride = 0;
st->hw_nblocksy = nblocksy;
st->hwbuf = svga_winsys_buffer_create(svga,
1,
0,
st->hw_nblocksy*st->base.stride);
while(!st->hwbuf && (st->hw_nblocksy /= 2)) {
st->hwbuf = svga_winsys_buffer_create(svga,
1,
0,
st->hw_nblocksy*st->base.stride);
}
if(!st->hwbuf)
goto no_hwbuf;
if(st->hw_nblocksy < nblocksy) {
/* We couldn't allocate a hardware buffer big enough for the transfer,
* so allocate regular malloc memory instead */
if (0) {
debug_printf("%s: failed to allocate %u KB of DMA, "
"splitting into %u x %u KB DMA transfers\n",
__FUNCTION__,
(nblocksy*st->base.stride + 1023)/1024,
(nblocksy + st->hw_nblocksy - 1)/st->hw_nblocksy,
(st->hw_nblocksy*st->base.stride + 1023)/1024);
}
st->swbuf = MALLOC(nblocksy*st->base.stride);
if(!st->swbuf)
goto no_swbuf;
}
if (usage & PIPE_TRANSFER_READ) {
SVGA3dSurfaceDMAFlags flags;
memset(&flags, 0, sizeof flags);
svga_transfer_dma(svga, st, SVGA3D_READ_HOST_VRAM, flags);
}
return &st->base;
no_swbuf:
sws->buffer_destroy(sws, st->hwbuf);
no_hwbuf:
FREE(st);
return NULL;
}
/* XXX: Still implementing this as if it was a screen function, but
* can now modify it to queue transfers on the context.
*/
static void *
svga_texture_transfer_map( struct pipe_context *pipe,
struct pipe_transfer *transfer )
{
struct svga_screen *ss = svga_screen(pipe->screen);
struct svga_winsys_screen *sws = ss->sws;
struct svga_transfer *st = svga_transfer(transfer);
if(st->swbuf)
return st->swbuf;
else
/* The wait for read transfers already happened when svga_transfer_dma
* was called. */
return sws->buffer_map(sws, st->hwbuf, transfer->usage);
}
/* XXX: Still implementing this as if it was a screen function, but
* can now modify it to queue transfers on the context.
*/
static void
svga_texture_transfer_unmap(struct pipe_context *pipe,
struct pipe_transfer *transfer)
{
struct svga_screen *ss = svga_screen(pipe->screen);
struct svga_winsys_screen *sws = ss->sws;
struct svga_transfer *st = svga_transfer(transfer);
if(!st->swbuf)
sws->buffer_unmap(sws, st->hwbuf);
}
static void
svga_texture_transfer_destroy(struct pipe_context *pipe,
struct pipe_transfer *transfer)
{
struct svga_context *svga = svga_context(pipe);
struct svga_texture *tex = svga_texture(transfer->resource);
struct svga_screen *ss = svga_screen(pipe->screen);
struct svga_winsys_screen *sws = ss->sws;
struct svga_transfer *st = svga_transfer(transfer);
if (st->base.usage & PIPE_TRANSFER_WRITE) {
SVGA3dSurfaceDMAFlags flags;
memset(&flags, 0, sizeof flags);
if (transfer->usage & PIPE_TRANSFER_DISCARD_WHOLE_RESOURCE) {
flags.discard = TRUE;
}
if (transfer->usage & PIPE_TRANSFER_UNSYNCHRONIZED) {
flags.unsynchronized = TRUE;
}
svga_transfer_dma(svga, st, SVGA3D_WRITE_HOST_VRAM, flags);
ss->texture_timestamp++;
tex->view_age[transfer->level] = ++(tex->age);
if (transfer->resource->target == PIPE_TEXTURE_CUBE)
tex->defined[transfer->box.z][transfer->level] = TRUE;
else
tex->defined[0][transfer->level] = TRUE;
}
pipe_resource_reference(&st->base.resource, NULL);
FREE(st->swbuf);
sws->buffer_destroy(sws, st->hwbuf);
FREE(st);
}
struct u_resource_vtbl svga_texture_vtbl =
{
svga_texture_get_handle, /* get_handle */
svga_texture_destroy, /* resource_destroy */
svga_texture_get_transfer, /* get_transfer */
svga_texture_transfer_destroy, /* transfer_destroy */
svga_texture_transfer_map, /* transfer_map */
u_default_transfer_flush_region, /* transfer_flush_region */
svga_texture_transfer_unmap, /* transfer_unmap */
u_default_transfer_inline_write /* transfer_inline_write */
};
struct pipe_resource *
svga_texture_create(struct pipe_screen *screen,
const struct pipe_resource *template)
{
struct svga_screen *svgascreen = svga_screen(screen);
struct svga_texture *tex = CALLOC_STRUCT(svga_texture);
if (!tex)
goto error1;
tex->b.b = *template;
tex->b.vtbl = &svga_texture_vtbl;
pipe_reference_init(&tex->b.b.reference, 1);
tex->b.b.screen = screen;
assert(template->last_level < SVGA_MAX_TEXTURE_LEVELS);
if(template->last_level >= SVGA_MAX_TEXTURE_LEVELS)
goto error2;
tex->key.flags = 0;
tex->key.size.width = template->width0;
tex->key.size.height = template->height0;
tex->key.size.depth = template->depth0;
if(template->target == PIPE_TEXTURE_CUBE) {
tex->key.flags |= SVGA3D_SURFACE_CUBEMAP;
tex->key.numFaces = 6;
}
else {
tex->key.numFaces = 1;
}
tex->key.cachable = 1;
if (template->bind & PIPE_BIND_SAMPLER_VIEW)
tex->key.flags |= SVGA3D_SURFACE_HINT_TEXTURE;
if (template->bind & PIPE_BIND_DISPLAY_TARGET) {
tex->key.cachable = 0;
}
if (template->bind & PIPE_BIND_SHARED) {
tex->key.cachable = 0;
}
if (template->bind & (PIPE_BIND_SCANOUT |
PIPE_BIND_CURSOR)) {
tex->key.flags |= SVGA3D_SURFACE_HINT_SCANOUT;
tex->key.cachable = 0;
}
/*
* Note: Previously we never passed the
* SVGA3D_SURFACE_HINT_RENDERTARGET hint. Mesa cannot
* know beforehand whether a texture will be used as a rendertarget or not
* and it always requests PIPE_BIND_RENDER_TARGET, therefore
* passing the SVGA3D_SURFACE_HINT_RENDERTARGET here defeats its purpose.
*
* However, this was changed since other state trackers
* (XA for example) uses it accurately and certain device versions
* relies on it in certain situations to render correctly.
*/
if((template->bind & PIPE_BIND_RENDER_TARGET) &&
!util_format_is_s3tc(template->format))
tex->key.flags |= SVGA3D_SURFACE_HINT_RENDERTARGET;
if(template->bind & PIPE_BIND_DEPTH_STENCIL)
tex->key.flags |= SVGA3D_SURFACE_HINT_DEPTHSTENCIL;
tex->key.numMipLevels = template->last_level + 1;
tex->key.format = svga_translate_format(svgascreen, template->format, template->bind);
if(tex->key.format == SVGA3D_FORMAT_INVALID)
goto error2;
SVGA_DBG(DEBUG_DMA, "surface_create for texture\n", tex->handle);
tex->handle = svga_screen_surface_create(svgascreen, &tex->key);
if (tex->handle)
SVGA_DBG(DEBUG_DMA, " --> got sid %p (texture)\n", tex->handle);
debug_reference(&tex->b.b.reference,
(debug_reference_descriptor)debug_describe_resource, 0);
return &tex->b.b;
error2:
FREE(tex);
error1:
return NULL;
}
struct pipe_resource *
svga_texture_from_handle(struct pipe_screen *screen,
const struct pipe_resource *template,
struct winsys_handle *whandle)
{
struct svga_winsys_screen *sws = svga_winsys_screen(screen);
struct svga_winsys_surface *srf;
struct svga_texture *tex;
enum SVGA3dSurfaceFormat format = 0;
assert(screen);
/* Only supports one type */
if ((template->target != PIPE_TEXTURE_2D &&
template->target != PIPE_TEXTURE_RECT) ||
template->last_level != 0 ||
template->depth0 != 1) {
return NULL;
}
srf = sws->surface_from_handle(sws, whandle, &format);
if (!srf)
return NULL;
if (svga_translate_format(svga_screen(screen), template->format, template->bind) != format) {
unsigned f1 = svga_translate_format(svga_screen(screen), template->format, template->bind);
unsigned f2 = format;
/* It's okay for XRGB and ARGB or depth with/out stencil to get mixed up */
if ( !( (f1 == SVGA3D_X8R8G8B8 && f2 == SVGA3D_A8R8G8B8) ||
(f1 == SVGA3D_A8R8G8B8 && f2 == SVGA3D_X8R8G8B8) ||
(f1 == SVGA3D_Z_D24X8 && f2 == SVGA3D_Z_D24S8) ||
(f1 == SVGA3D_Z_DF24 && f2 == SVGA3D_Z_D24S8_INT) ) ) {
debug_printf("%s wrong format %u != %u\n", __FUNCTION__, f1, f2);
return NULL;
}
}
tex = CALLOC_STRUCT(svga_texture);
if (!tex)
return NULL;
tex->b.b = *template;
tex->b.vtbl = &svga_texture_vtbl;
pipe_reference_init(&tex->b.b.reference, 1);
tex->b.b.screen = screen;
if (format == SVGA3D_X8R8G8B8)
tex->b.b.format = PIPE_FORMAT_B8G8R8X8_UNORM;
else if (format == SVGA3D_A8R8G8B8)
tex->b.b.format = PIPE_FORMAT_B8G8R8A8_UNORM;
else {
/* ?? */
}
SVGA_DBG(DEBUG_DMA, "wrap surface sid %p\n", srf);
tex->key.cachable = 0;
tex->handle = srf;
return &tex->b.b;
}
|