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
|
/*
* Copyright © 2012 Intel Corporation
*
* 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 (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 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 <errno.h>
#include "intel_batchbuffer.h"
#include "intel_fbo.h"
#include "brw_blorp.h"
#include "brw_compiler.h"
#include "brw_nir.h"
#include "brw_state.h"
#define FILE_DEBUG_FLAG DEBUG_BLORP
static void
apply_gen6_stencil_hiz_offset(struct isl_surf *surf,
struct intel_mipmap_tree *mt,
uint32_t lod,
uint32_t *offset)
{
assert(mt->array_layout == ALL_SLICES_AT_EACH_LOD);
if (mt->format == MESA_FORMAT_S_UINT8) {
/* Note: we can't compute the stencil offset using
* intel_miptree_get_aligned_offset(), because the miptree
* claims that the region is untiled even though it's W tiled.
*/
*offset = mt->level[lod].level_y * mt->pitch +
mt->level[lod].level_x * 64;
} else {
*offset = intel_miptree_get_aligned_offset(mt,
mt->level[lod].level_x,
mt->level[lod].level_y,
false);
}
surf->logical_level0_px.width = minify(surf->logical_level0_px.width, lod);
surf->logical_level0_px.height = minify(surf->logical_level0_px.height, lod);
surf->phys_level0_sa.width = minify(surf->phys_level0_sa.width, lod);
surf->phys_level0_sa.height = minify(surf->phys_level0_sa.height, lod);
surf->levels = 1;
surf->array_pitch_el_rows =
ALIGN(surf->phys_level0_sa.height, surf->image_alignment_el.height);
}
void
brw_blorp_surf_for_miptree(struct brw_context *brw,
struct brw_blorp_surf *surf,
struct intel_mipmap_tree *mt,
bool is_render_target,
unsigned *level,
struct isl_surf tmp_surfs[2])
{
intel_miptree_get_isl_surf(brw, mt, &tmp_surfs[0]);
surf->surf = &tmp_surfs[0];
surf->bo = mt->bo;
surf->offset = mt->offset;
if (brw->gen == 6 && mt->format == MESA_FORMAT_S_UINT8 &&
mt->array_layout == ALL_SLICES_AT_EACH_LOD) {
/* Sandy bridge stencil and HiZ use this ALL_SLICES_AT_EACH_LOD hack in
* order to allow for layered rendering. The hack makes each LOD of the
* stencil or HiZ buffer a single tightly packed array surface at some
* offset into the surface. Since ISL doesn't know how to deal with the
* crazy ALL_SLICES_AT_EACH_LOD layout and since we have to do a manual
* offset of it anyway, we might as well do the offset here and keep the
* hacks inside the i965 driver.
*
* See also gen6_depth_stencil_state.c
*/
uint32_t offset;
apply_gen6_stencil_hiz_offset(&tmp_surfs[0], mt, *level, &offset);
surf->offset += offset;
*level = 0;
}
struct isl_surf *aux_surf = &tmp_surfs[1];
intel_miptree_get_aux_isl_surf(brw, mt, aux_surf, &surf->aux_usage);
/* For textures that are in the RESOLVED state, we ignore the MCS */
if (mt->mcs_mt && !is_render_target &&
mt->fast_clear_state == INTEL_FAST_CLEAR_STATE_RESOLVED)
surf->aux_usage = ISL_AUX_USAGE_NONE;
if (surf->aux_usage != ISL_AUX_USAGE_NONE) {
/* We only really need a clear color if we also have an auxiliary
* surface. Without one, it does nothing.
*/
surf->clear_color = intel_miptree_get_isl_clear_color(brw, mt);
surf->aux_surf = aux_surf;
if (mt->mcs_mt) {
surf->aux_bo = mt->mcs_mt->bo;
surf->aux_offset = mt->mcs_mt->offset;
} else {
assert(surf->aux_usage == ISL_AUX_USAGE_HIZ);
struct intel_mipmap_tree *hiz_mt = mt->hiz_buf->mt;
if (hiz_mt) {
surf->aux_bo = hiz_mt->bo;
if (brw->gen == 6 &&
hiz_mt->array_layout == ALL_SLICES_AT_EACH_LOD) {
/* gen6 requires the HiZ buffer to be manually offset to the
* right location. We could fixup the surf but it doesn't
* matter since most of those fields don't matter.
*/
apply_gen6_stencil_hiz_offset(aux_surf, hiz_mt, *level,
&surf->aux_offset);
} else {
surf->aux_offset = 0;
}
assert(hiz_mt->pitch == aux_surf->row_pitch);
} else {
surf->aux_bo = mt->hiz_buf->bo;
surf->aux_offset = 0;
}
}
} else {
surf->aux_bo = NULL;
surf->aux_offset = 0;
memset(&surf->clear_color, 0, sizeof(surf->clear_color));
}
assert((surf->aux_usage == ISL_AUX_USAGE_NONE) == (surf->aux_bo == NULL));
}
enum isl_format
brw_blorp_to_isl_format(struct brw_context *brw, mesa_format format,
bool is_render_target)
{
switch (format) {
case MESA_FORMAT_NONE:
return ISL_FORMAT_UNSUPPORTED;
case MESA_FORMAT_S_UINT8:
return ISL_FORMAT_R8_UINT;
case MESA_FORMAT_Z24_UNORM_X8_UINT:
return ISL_FORMAT_R24_UNORM_X8_TYPELESS;
case MESA_FORMAT_Z_FLOAT32:
return ISL_FORMAT_R32_FLOAT;
case MESA_FORMAT_Z_UNORM16:
return ISL_FORMAT_R16_UNORM;
default: {
if (is_render_target) {
assert(brw->format_supported_as_render_target[format]);
return brw->render_target_format[format];
} else {
return brw_format_for_mesa_format(format);
}
break;
}
}
}
void
brw_blorp_surface_info_init(struct brw_context *brw,
struct brw_blorp_surface_info *info,
const struct brw_blorp_surf *surf,
unsigned int level, unsigned int layer,
enum isl_format format, bool is_render_target)
{
/* Layer is a physical layer, so if this is a 2D multisample array texture
* using INTEL_MSAA_LAYOUT_UMS or INTEL_MSAA_LAYOUT_CMS, then it had better
* be a multiple of num_samples.
*/
unsigned layer_multiplier = 1;
if (surf->surf->msaa_layout == ISL_MSAA_LAYOUT_ARRAY) {
assert(layer % surf->surf->samples == 0);
layer_multiplier = surf->surf->samples;
}
if (format == ISL_FORMAT_UNSUPPORTED)
format = surf->surf->format;
if (format == ISL_FORMAT_R24_UNORM_X8_TYPELESS) {
/* Unfortunately, ISL_FORMAT_R24_UNORM_X8_TYPELESS it isn't supported as
* a render target, which would prevent us from blitting to 24-bit
* depth. The miptree consists of 32 bits per pixel, arranged as 24-bit
* depth values interleaved with 8 "don't care" bits. Since depth
* values don't require any blending, it doesn't matter how we interpret
* the bit pattern as long as we copy the right amount of data, so just
* map it as 8-bit BGRA.
*/
format = ISL_FORMAT_B8G8R8A8_UNORM;
} else if (surf->surf->usage & ISL_SURF_USAGE_STENCIL_BIT) {
assert(surf->surf->format == ISL_FORMAT_R8_UINT);
/* Prior to Broadwell, we can't render to R8_UINT */
if (brw->gen < 8)
format = ISL_FORMAT_R8_UNORM;
}
info->surf = *surf->surf;
info->bo = surf->bo;
info->offset = surf->offset;
info->aux_usage = surf->aux_usage;
if (info->aux_usage != ISL_AUX_USAGE_NONE) {
info->aux_surf = *surf->aux_surf;
info->aux_bo = surf->aux_bo;
info->aux_offset = surf->aux_offset;
}
info->clear_color = surf->clear_color;
info->view = (struct isl_view) {
.usage = is_render_target ? ISL_SURF_USAGE_RENDER_TARGET_BIT :
ISL_SURF_USAGE_TEXTURE_BIT,
.format = format,
.base_level = level,
.levels = 1,
.channel_select = {
ISL_CHANNEL_SELECT_RED,
ISL_CHANNEL_SELECT_GREEN,
ISL_CHANNEL_SELECT_BLUE,
ISL_CHANNEL_SELECT_ALPHA,
},
};
if (!is_render_target &&
(info->surf.dim == ISL_SURF_DIM_3D ||
info->surf.msaa_layout == ISL_MSAA_LAYOUT_ARRAY)) {
/* 3-D textures don't support base_array layer and neither do 2-D
* multisampled textures on IVB so we need to pass it through the
* sampler in those cases. These are also two cases where we are
* guaranteed that we won't be doing any funny surface hacks.
*/
info->view.base_array_layer = 0;
info->view.array_len = MAX2(info->surf.logical_level0_px.depth,
info->surf.logical_level0_px.array_len);
info->z_offset = layer / layer_multiplier;
} else {
info->view.base_array_layer = layer / layer_multiplier;
info->view.array_len = 1;
info->z_offset = 0;
}
}
void
brw_blorp_params_init(struct brw_blorp_params *params)
{
memset(params, 0, sizeof(*params));
params->hiz_op = GEN6_HIZ_OP_NONE;
params->fast_clear_op = 0;
params->num_draw_buffers = 1;
params->num_layers = 1;
}
void
brw_blorp_init_wm_prog_key(struct brw_wm_prog_key *wm_key)
{
memset(wm_key, 0, sizeof(*wm_key));
wm_key->nr_color_regions = 1;
for (int i = 0; i < MAX_SAMPLERS; i++)
wm_key->tex.swizzles[i] = SWIZZLE_XYZW;
}
static int
nir_uniform_type_size(const struct glsl_type *type)
{
/* Only very basic types are allowed */
assert(glsl_type_is_vector_or_scalar(type));
assert(glsl_get_bit_size(type) == 32);
return glsl_get_vector_elements(type) * 4;
}
const unsigned *
brw_blorp_compile_nir_shader(struct brw_context *brw, struct nir_shader *nir,
const struct brw_wm_prog_key *wm_key,
bool use_repclear,
struct brw_blorp_prog_data *prog_data,
unsigned *program_size)
{
const struct brw_compiler *compiler = brw->intelScreen->compiler;
void *mem_ctx = ralloc_context(NULL);
/* Calling brw_preprocess_nir and friends is destructive and, if cloning is
* enabled, may end up completely replacing the nir_shader. Therefore, we
* own it and might as well put it in our context for easy cleanup.
*/
ralloc_steal(mem_ctx, nir);
nir->options =
compiler->glsl_compiler_options[MESA_SHADER_FRAGMENT].NirOptions;
struct brw_wm_prog_data wm_prog_data;
memset(&wm_prog_data, 0, sizeof(wm_prog_data));
wm_prog_data.base.nr_params = 0;
wm_prog_data.base.param = NULL;
/* BLORP always just uses the first two binding table entries */
wm_prog_data.binding_table.render_target_start = 0;
wm_prog_data.base.binding_table.texture_start = 1;
nir = brw_preprocess_nir(compiler, nir);
nir_remove_dead_variables(nir, nir_var_shader_in);
nir_shader_gather_info(nir, nir_shader_get_entrypoint(nir)->impl);
/* Uniforms are required to be lowered before going into compile_fs. For
* BLORP, we'll assume that whoever builds the shader sets the location
* they want so we just need to lower them and figure out how many we have
* in total.
*/
nir->num_uniforms = 0;
nir_foreach_variable(var, &nir->uniforms) {
var->data.driver_location = var->data.location;
unsigned end = var->data.location + nir_uniform_type_size(var->type);
nir->num_uniforms = MAX2(nir->num_uniforms, end);
}
nir_lower_io(nir, nir_var_uniform, nir_uniform_type_size);
const unsigned *program =
brw_compile_fs(compiler, brw, mem_ctx, wm_key, &wm_prog_data, nir,
NULL, -1, -1, false, use_repclear, program_size, NULL);
/* Copy the relavent bits of wm_prog_data over into the blorp prog data */
prog_data->dispatch_8 = wm_prog_data.dispatch_8;
prog_data->dispatch_16 = wm_prog_data.dispatch_16;
prog_data->first_curbe_grf_0 = wm_prog_data.base.dispatch_grf_start_reg;
prog_data->first_curbe_grf_2 = wm_prog_data.dispatch_grf_start_reg_2;
prog_data->ksp_offset_2 = wm_prog_data.prog_offset_2;
prog_data->persample_msaa_dispatch = wm_prog_data.persample_dispatch;
prog_data->flat_inputs = wm_prog_data.flat_inputs;
prog_data->num_varying_inputs = wm_prog_data.num_varying_inputs;
prog_data->inputs_read = nir->info.inputs_read;
assert(wm_prog_data.base.nr_params == 0);
return program;
}
struct surface_state_info {
unsigned num_dwords;
unsigned ss_align; /* Required alignment of RENDER_SURFACE_STATE in bytes */
unsigned reloc_dw;
unsigned aux_reloc_dw;
unsigned tex_mocs;
unsigned rb_mocs;
};
static const struct surface_state_info surface_state_infos[] = {
[6] = {6, 32, 1, 0},
[7] = {8, 32, 1, 6, GEN7_MOCS_L3, GEN7_MOCS_L3},
[8] = {13, 64, 8, 10, BDW_MOCS_WB, BDW_MOCS_PTE},
[9] = {16, 64, 8, 10, SKL_MOCS_WB, SKL_MOCS_PTE},
};
uint32_t
brw_blorp_emit_surface_state(struct brw_context *brw,
const struct brw_blorp_surface_info *surface,
uint32_t read_domains, uint32_t write_domain,
bool is_render_target)
{
const struct surface_state_info ss_info = surface_state_infos[brw->gen];
struct isl_surf surf = surface->surf;
if (surf.dim == ISL_SURF_DIM_1D &&
surf.dim_layout == ISL_DIM_LAYOUT_GEN4_2D) {
assert(surf.logical_level0_px.height == 1);
surf.dim = ISL_SURF_DIM_2D;
}
/* Blorp doesn't support HiZ in any of the blit or slow-clear paths */
enum isl_aux_usage aux_usage = surface->aux_usage;
if (aux_usage == ISL_AUX_USAGE_HIZ)
aux_usage = ISL_AUX_USAGE_NONE;
uint32_t surf_offset;
uint32_t *dw = brw_state_batch(brw, AUB_TRACE_SURFACE_STATE,
ss_info.num_dwords * 4, ss_info.ss_align,
&surf_offset);
const uint32_t mocs = is_render_target ? ss_info.rb_mocs : ss_info.tex_mocs;
uint64_t aux_bo_offset = surface->aux_bo ? surface->aux_bo->offset64 : 0;
isl_surf_fill_state(&brw->isl_dev, dw, .surf = &surf, .view = &surface->view,
.address = surface->bo->offset64 + surface->offset,
.aux_surf = &surface->aux_surf, .aux_usage = aux_usage,
.aux_address = aux_bo_offset + surface->aux_offset,
.mocs = mocs, .clear_color = surface->clear_color,
.x_offset_sa = surface->tile_x_sa,
.y_offset_sa = surface->tile_y_sa);
/* Emit relocation to surface contents */
drm_intel_bo_emit_reloc(brw->batch.bo,
surf_offset + ss_info.reloc_dw * 4,
surface->bo,
dw[ss_info.reloc_dw] - surface->bo->offset64,
read_domains, write_domain);
if (aux_usage != ISL_AUX_USAGE_NONE) {
/* On gen7 and prior, the bottom 12 bits of the MCS base address are
* used to store other information. This should be ok, however, because
* surface buffer addresses are always 4K page alinged.
*/
assert((surface->aux_offset & 0xfff) == 0);
drm_intel_bo_emit_reloc(brw->batch.bo,
surf_offset + ss_info.aux_reloc_dw * 4,
surface->aux_bo,
dw[ss_info.aux_reloc_dw] & 0xfff,
read_domains, write_domain);
}
return surf_offset;
}
/**
* Perform a HiZ or depth resolve operation.
*
* For an overview of HiZ ops, see the following sections of the Sandy Bridge
* PRM, Volume 1, Part 2:
* - 7.5.3.1 Depth Buffer Clear
* - 7.5.3.2 Depth Buffer Resolve
* - 7.5.3.3 Hierarchical Depth Buffer Resolve
*/
void
intel_hiz_exec(struct brw_context *brw, struct intel_mipmap_tree *mt,
unsigned int level, unsigned int layer, enum gen6_hiz_op op)
{
const char *opname = NULL;
switch (op) {
case GEN6_HIZ_OP_DEPTH_RESOLVE:
opname = "depth resolve";
break;
case GEN6_HIZ_OP_HIZ_RESOLVE:
opname = "hiz ambiguate";
break;
case GEN6_HIZ_OP_DEPTH_CLEAR:
opname = "depth clear";
break;
case GEN6_HIZ_OP_NONE:
opname = "noop?";
break;
}
DBG("%s %s to mt %p level %d layer %d\n",
__func__, opname, mt, level, layer);
if (brw->gen >= 8) {
gen8_hiz_exec(brw, mt, level, layer, op);
} else {
gen6_blorp_hiz_exec(brw, mt, level, layer, op);
}
}
void
brw_blorp_exec(struct brw_context *brw, const struct brw_blorp_params *params)
{
struct gl_context *ctx = &brw->ctx;
const uint32_t estimated_max_batch_usage = brw->gen >= 8 ? 1800 : 1500;
bool check_aperture_failed_once = false;
/* Flush the sampler and render caches. We definitely need to flush the
* sampler cache so that we get updated contents from the render cache for
* the glBlitFramebuffer() source. Also, we are sometimes warned in the
* docs to flush the cache between reinterpretations of the same surface
* data with different formats, which blorp does for stencil and depth
* data.
*/
brw_emit_mi_flush(brw);
brw_select_pipeline(brw, BRW_RENDER_PIPELINE);
retry:
intel_batchbuffer_require_space(brw, estimated_max_batch_usage, RENDER_RING);
intel_batchbuffer_save_state(brw);
drm_intel_bo *saved_bo = brw->batch.bo;
uint32_t saved_used = USED_BATCH(brw->batch);
uint32_t saved_state_batch_offset = brw->batch.state_batch_offset;
switch (brw->gen) {
case 6:
gen6_blorp_exec(brw, params);
break;
case 7:
gen7_blorp_exec(brw, params);
break;
case 8:
case 9:
gen8_blorp_exec(brw, params);
break;
default:
/* BLORP is not supported before Gen6. */
unreachable("not reached");
}
/* Make sure we didn't wrap the batch unintentionally, and make sure we
* reserved enough space that a wrap will never happen.
*/
assert(brw->batch.bo == saved_bo);
assert((USED_BATCH(brw->batch) - saved_used) * 4 +
(saved_state_batch_offset - brw->batch.state_batch_offset) <
estimated_max_batch_usage);
/* Shut up compiler warnings on release build */
(void)saved_bo;
(void)saved_used;
(void)saved_state_batch_offset;
/* Check if the blorp op we just did would make our batch likely to fail to
* map all the BOs into the GPU at batch exec time later. If so, flush the
* batch and try again with nothing else in the batch.
*/
if (dri_bufmgr_check_aperture_space(&brw->batch.bo, 1)) {
if (!check_aperture_failed_once) {
check_aperture_failed_once = true;
intel_batchbuffer_reset_to_saved(brw);
intel_batchbuffer_flush(brw);
goto retry;
} else {
int ret = intel_batchbuffer_flush(brw);
WARN_ONCE(ret == -ENOSPC,
"i965: blorp emit exceeded available aperture space\n");
}
}
if (unlikely(brw->always_flush_batch))
intel_batchbuffer_flush(brw);
/* We've smashed all state compared to what the normal 3D pipeline
* rendering tracks for GL.
*/
brw->ctx.NewDriverState |= BRW_NEW_BLORP;
brw->no_depth_or_stencil = false;
brw->ib.type = -1;
/* Flush the sampler cache so any texturing from the destination is
* coherent.
*/
brw_emit_mi_flush(brw);
}
static void
blorp_gen6_hiz_op(struct brw_context *brw, struct brw_blorp_surf *surf,
unsigned level, unsigned layer, enum gen6_hiz_op op)
{
struct brw_blorp_params params;
brw_blorp_params_init(¶ms);
params.hiz_op = op;
brw_blorp_surface_info_init(brw, ¶ms.depth, surf, level, layer,
surf->surf->format, true);
/* Align the rectangle primitive to 8x4 pixels.
*
* During fast depth clears, the emitted rectangle primitive must be
* aligned to 8x4 pixels. From the Ivybridge PRM, Vol 2 Part 1 Section
* 11.5.3.1 Depth Buffer Clear (and the matching section in the Sandybridge
* PRM):
* If Number of Multisamples is NUMSAMPLES_1, the rectangle must be
* aligned to an 8x4 pixel block relative to the upper left corner
* of the depth buffer [...]
*
* For hiz resolves, the rectangle must also be 8x4 aligned. Item
* WaHizAmbiguate8x4Aligned from the Haswell workarounds page and the
* Ivybridge simulator require the alignment.
*
* To be safe, let's just align the rect for all hiz operations and all
* hardware generations.
*
* However, for some miptree slices of a Z24 texture, emitting an 8x4
* aligned rectangle that covers the slice may clobber adjacent slices if
* we strictly adhered to the texture alignments specified in the PRM. The
* Ivybridge PRM, Section "Alignment Unit Size", states that
* SURFACE_STATE.Surface_Horizontal_Alignment should be 4 for Z24 surfaces,
* not 8. But commit 1f112cc increased the alignment from 4 to 8, which
* prevents the clobbering.
*/
params.x1 = minify(params.depth.surf.logical_level0_px.width,
params.depth.view.base_level);
params.y1 = minify(params.depth.surf.logical_level0_px.height,
params.depth.view.base_level);
params.x1 = ALIGN(params.x1, 8);
params.y1 = ALIGN(params.y1, 4);
if (params.depth.view.base_level == 0) {
/* TODO: What about MSAA? */
params.depth.surf.logical_level0_px.width = params.x1;
params.depth.surf.logical_level0_px.height = params.y1;
}
params.dst.surf.samples = params.depth.surf.samples;
params.dst.surf.logical_level0_px = params.depth.surf.logical_level0_px;
switch (surf->surf->format) {
case ISL_FORMAT_R16_UNORM:
params.depth_format = BRW_DEPTHFORMAT_D16_UNORM;
break;
case ISL_FORMAT_R32_FLOAT:
params.depth_format = BRW_DEPTHFORMAT_D32_FLOAT;
break;
case ISL_FORMAT_R24_UNORM_X8_TYPELESS:
params.depth_format = BRW_DEPTHFORMAT_D24_UNORM_X8_UINT;
break;
default:
unreachable("not reached");
}
brw_blorp_exec(brw, ¶ms);
}
void
gen6_blorp_hiz_exec(struct brw_context *brw, struct intel_mipmap_tree *mt,
unsigned int level, unsigned int layer, enum gen6_hiz_op op)
{
intel_miptree_check_level_layer(mt, level, layer);
intel_miptree_used_for_rendering(mt);
assert(intel_miptree_level_has_hiz(mt, level));
struct isl_surf isl_tmp[2];
struct brw_blorp_surf surf;
brw_blorp_surf_for_miptree(brw, &surf, mt, true, &level, isl_tmp);
blorp_gen6_hiz_op(brw, &surf, level, layer, op);
}
|