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
|
/*
* Copyright © 2017 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 "anv_nir.h"
#include "anv_private.h"
#include "nir/nir.h"
#include "nir/nir_builder.h"
struct ycbcr_state {
nir_builder *builder;
nir_ssa_def *image_size;
nir_tex_instr *origin_tex;
nir_deref_instr *tex_deref;
struct anv_ycbcr_conversion *conversion;
};
static nir_ssa_def *
y_range(nir_builder *b,
nir_ssa_def *y_channel,
int bpc,
VkSamplerYcbcrRange range)
{
switch (range) {
case VK_SAMPLER_YCBCR_RANGE_ITU_FULL:
return y_channel;
case VK_SAMPLER_YCBCR_RANGE_ITU_NARROW:
return nir_fmul(b,
nir_fadd(b,
nir_fmul(b, y_channel,
nir_imm_float(b, pow(2, bpc) - 1)),
nir_imm_float(b, -16.0f * pow(2, bpc - 8))),
nir_frcp(b, nir_imm_float(b, 219.0f * pow(2, bpc - 8))));
default:
unreachable("missing Ycbcr range");
return NULL;
}
}
static nir_ssa_def *
chroma_range(nir_builder *b,
nir_ssa_def *chroma_channel,
int bpc,
VkSamplerYcbcrRange range)
{
switch (range) {
case VK_SAMPLER_YCBCR_RANGE_ITU_FULL:
return nir_fadd(b, chroma_channel,
nir_imm_float(b, -pow(2, bpc - 1) / (pow(2, bpc) - 1.0f)));
case VK_SAMPLER_YCBCR_RANGE_ITU_NARROW:
return nir_fmul(b,
nir_fadd(b,
nir_fmul(b, chroma_channel,
nir_imm_float(b, pow(2, bpc) - 1)),
nir_imm_float(b, -128.0f * pow(2, bpc - 8))),
nir_frcp(b, nir_imm_float(b, 224.0f * pow(2, bpc - 8))));
default:
unreachable("missing Ycbcr range");
return NULL;
}
}
typedef struct nir_const_value_3_4 {
nir_const_value v[3][4];
} nir_const_value_3_4;
static const nir_const_value_3_4 *
ycbcr_model_to_rgb_matrix(VkSamplerYcbcrModelConversion model)
{
switch (model) {
case VK_SAMPLER_YCBCR_MODEL_CONVERSION_YCBCR_601: {
static const nir_const_value_3_4 bt601 = { {
{ { .f32 = 1.402f }, { .f32 = 1.0f }, { .f32 = 0.0f }, { .f32 = 0.0f } },
{ { .f32 = -0.714136286201022f }, { .f32 = 1.0f }, { .f32 = -0.344136286201022f }, { .f32 = 0.0f } },
{ { .f32 = 0.0f }, { .f32 = 1.0f }, { .f32 = 1.772f }, { .f32 = 0.0f } },
} };
return &bt601;
}
case VK_SAMPLER_YCBCR_MODEL_CONVERSION_YCBCR_709: {
static const nir_const_value_3_4 bt709 = { {
{ { .f32 = 1.5748031496063f }, { .f32 = 1.0f }, { .f32 = 0.0f }, { .f32 = 0.0f } },
{ { .f32 = -0.468125209181067f }, { .f32 = 1.0f }, { .f32 = -0.187327487470334f }, { .f32 = 0.0f } },
{ { .f32 = 0.0f }, { .f32 = 1.0f }, { .f32 = 1.85563184264242f }, { .f32 = 0.0f } },
} };
return &bt709;
}
case VK_SAMPLER_YCBCR_MODEL_CONVERSION_YCBCR_2020: {
static const nir_const_value_3_4 bt2020 = { {
{ { .f32 = 1.4746f }, { .f32 = 1.0f }, { .f32 = 0.0f }, { .f32 = 0.0f } },
{ { .f32 = -0.571353126843658f }, { .f32 = 1.0f }, { .f32 = -0.164553126843658f }, { .f32 = 0.0f } },
{ { .f32 = 0.0f }, { .f32 = 1.0f }, { .f32 = 1.8814f }, { .f32 = 0.0f } },
} };
return &bt2020;
}
default:
unreachable("missing Ycbcr model");
return NULL;
}
}
static nir_ssa_def *
convert_ycbcr(struct ycbcr_state *state,
nir_ssa_def *raw_channels,
uint32_t *bpcs)
{
nir_builder *b = state->builder;
struct anv_ycbcr_conversion *conversion = state->conversion;
nir_ssa_def *expanded_channels =
nir_vec4(b,
chroma_range(b, nir_channel(b, raw_channels, 0),
bpcs[0], conversion->ycbcr_range),
y_range(b, nir_channel(b, raw_channels, 1),
bpcs[1], conversion->ycbcr_range),
chroma_range(b, nir_channel(b, raw_channels, 2),
bpcs[2], conversion->ycbcr_range),
nir_imm_float(b, 1.0f));
if (conversion->ycbcr_model == VK_SAMPLER_YCBCR_MODEL_CONVERSION_YCBCR_IDENTITY)
return expanded_channels;
const nir_const_value_3_4 *conversion_matrix =
ycbcr_model_to_rgb_matrix(conversion->ycbcr_model);
nir_ssa_def *converted_channels[] = {
nir_fdot4(b, expanded_channels, nir_build_imm(b, 4, 32, conversion_matrix->v[0])),
nir_fdot4(b, expanded_channels, nir_build_imm(b, 4, 32, conversion_matrix->v[1])),
nir_fdot4(b, expanded_channels, nir_build_imm(b, 4, 32, conversion_matrix->v[2]))
};
return nir_vec4(b,
converted_channels[0], converted_channels[1],
converted_channels[2], nir_imm_float(b, 1.0f));
}
/* TODO: we should probably replace this with a push constant/uniform. */
static nir_ssa_def *
get_texture_size(struct ycbcr_state *state, nir_deref_instr *texture)
{
if (state->image_size)
return state->image_size;
nir_builder *b = state->builder;
const struct glsl_type *type = texture->type;
nir_tex_instr *tex = nir_tex_instr_create(b->shader, 1);
tex->op = nir_texop_txs;
tex->sampler_dim = glsl_get_sampler_dim(type);
tex->is_array = glsl_sampler_type_is_array(type);
tex->is_shadow = glsl_sampler_type_is_shadow(type);
tex->dest_type = nir_type_int;
tex->src[0].src_type = nir_tex_src_texture_deref;
tex->src[0].src = nir_src_for_ssa(&texture->dest.ssa);
nir_ssa_dest_init(&tex->instr, &tex->dest,
nir_tex_instr_dest_size(tex), 32, NULL);
nir_builder_instr_insert(b, &tex->instr);
state->image_size = nir_i2f32(b, &tex->dest.ssa);
return state->image_size;
}
static nir_ssa_def *
implicit_downsampled_coord(nir_builder *b,
nir_ssa_def *value,
nir_ssa_def *max_value,
int div_scale)
{
return nir_fadd(b,
value,
nir_fdiv(b,
nir_imm_float(b, 1.0f),
nir_fmul(b,
nir_imm_float(b, div_scale),
max_value)));
}
static nir_ssa_def *
implicit_downsampled_coords(struct ycbcr_state *state,
nir_ssa_def *old_coords,
const struct anv_format_plane *plane_format)
{
nir_builder *b = state->builder;
struct anv_ycbcr_conversion *conversion = state->conversion;
nir_ssa_def *image_size = get_texture_size(state, state->tex_deref);
nir_ssa_def *comp[4] = { NULL, };
int c;
for (c = 0; c < ARRAY_SIZE(conversion->chroma_offsets); c++) {
if (plane_format->denominator_scales[c] > 1 &&
conversion->chroma_offsets[c] == VK_CHROMA_LOCATION_COSITED_EVEN) {
comp[c] = implicit_downsampled_coord(b,
nir_channel(b, old_coords, c),
nir_channel(b, image_size, c),
plane_format->denominator_scales[c]);
} else {
comp[c] = nir_channel(b, old_coords, c);
}
}
/* Leave other coordinates untouched */
for (; c < old_coords->num_components; c++)
comp[c] = nir_channel(b, old_coords, c);
return nir_vec(b, comp, old_coords->num_components);
}
static nir_ssa_def *
create_plane_tex_instr_implicit(struct ycbcr_state *state,
uint32_t plane)
{
nir_builder *b = state->builder;
struct anv_ycbcr_conversion *conversion = state->conversion;
const struct anv_format_plane *plane_format =
&conversion->format->planes[plane];
nir_tex_instr *old_tex = state->origin_tex;
nir_tex_instr *tex = nir_tex_instr_create(b->shader, old_tex->num_srcs + 1);
for (uint32_t i = 0; i < old_tex->num_srcs; i++) {
tex->src[i].src_type = old_tex->src[i].src_type;
switch (old_tex->src[i].src_type) {
case nir_tex_src_coord:
if (plane_format->has_chroma && conversion->chroma_reconstruction) {
assert(old_tex->src[i].src.is_ssa);
tex->src[i].src =
nir_src_for_ssa(implicit_downsampled_coords(state,
old_tex->src[i].src.ssa,
plane_format));
break;
}
/* fall through */
default:
nir_src_copy(&tex->src[i].src, &old_tex->src[i].src, tex);
break;
}
}
tex->src[tex->num_srcs - 1].src = nir_src_for_ssa(nir_imm_int(b, plane));
tex->src[tex->num_srcs - 1].src_type = nir_tex_src_plane;
tex->sampler_dim = old_tex->sampler_dim;
tex->dest_type = old_tex->dest_type;
tex->op = old_tex->op;
tex->coord_components = old_tex->coord_components;
tex->is_new_style_shadow = old_tex->is_new_style_shadow;
tex->component = old_tex->component;
tex->texture_index = old_tex->texture_index;
tex->texture_array_size = old_tex->texture_array_size;
tex->sampler_index = old_tex->sampler_index;
tex->is_array = old_tex->is_array;
nir_ssa_dest_init(&tex->instr, &tex->dest,
old_tex->dest.ssa.num_components,
nir_dest_bit_size(old_tex->dest), NULL);
nir_builder_instr_insert(b, &tex->instr);
return &tex->dest.ssa;
}
static unsigned
channel_to_component(enum isl_channel_select channel)
{
switch (channel) {
case ISL_CHANNEL_SELECT_RED:
return 0;
case ISL_CHANNEL_SELECT_GREEN:
return 1;
case ISL_CHANNEL_SELECT_BLUE:
return 2;
case ISL_CHANNEL_SELECT_ALPHA:
return 3;
default:
unreachable("invalid channel");
return 0;
}
}
static enum isl_channel_select
swizzle_channel(struct isl_swizzle swizzle, unsigned channel)
{
switch (channel) {
case 0:
return swizzle.r;
case 1:
return swizzle.g;
case 2:
return swizzle.b;
case 3:
return swizzle.a;
default:
unreachable("invalid channel");
return 0;
}
}
static bool
try_lower_tex_ycbcr(struct anv_pipeline_layout *layout,
nir_builder *builder,
nir_tex_instr *tex)
{
int deref_src_idx = nir_tex_instr_src_index(tex, nir_tex_src_texture_deref);
assert(deref_src_idx >= 0);
nir_deref_instr *deref = nir_src_as_deref(tex->src[deref_src_idx].src);
nir_variable *var = nir_deref_instr_get_variable(deref);
const struct anv_descriptor_set_layout *set_layout =
layout->set[var->data.descriptor_set].layout;
const struct anv_descriptor_set_binding_layout *binding =
&set_layout->binding[var->data.binding];
/* For the following instructions, we don't apply any change and let the
* instruction apply to the first plane.
*/
if (tex->op == nir_texop_txs ||
tex->op == nir_texop_query_levels ||
tex->op == nir_texop_lod)
return false;
if (binding->immutable_samplers == NULL)
return false;
assert(tex->texture_index == 0);
unsigned array_index = 0;
if (deref->deref_type != nir_deref_type_var) {
assert(deref->deref_type == nir_deref_type_array);
if (!nir_src_is_const(deref->arr.index))
return false;
array_index = nir_src_as_uint(deref->arr.index);
array_index = MIN2(array_index, binding->array_size - 1);
}
const struct anv_sampler *sampler = binding->immutable_samplers[array_index];
if (sampler->conversion == NULL)
return false;
struct ycbcr_state state = {
.builder = builder,
.origin_tex = tex,
.tex_deref = deref,
.conversion = sampler->conversion,
};
builder->cursor = nir_before_instr(&tex->instr);
const struct anv_format *format = state.conversion->format;
const struct isl_format_layout *y_isl_layout = NULL;
for (uint32_t p = 0; p < format->n_planes; p++) {
if (!format->planes[p].has_chroma)
y_isl_layout = isl_format_get_layout(format->planes[p].isl_format);
}
assert(y_isl_layout != NULL);
uint8_t y_bpc = y_isl_layout->channels_array[0].bits;
/* |ycbcr_comp| holds components in the order : Cr-Y-Cb */
nir_ssa_def *zero = nir_imm_float(builder, 0.0f);
nir_ssa_def *one = nir_imm_float(builder, 1.0f);
/* Use extra 2 channels for following swizzle */
nir_ssa_def *ycbcr_comp[5] = { zero, zero, zero, one, zero };
uint8_t ycbcr_bpcs[5];
memset(ycbcr_bpcs, y_bpc, sizeof(ycbcr_bpcs));
/* Go through all the planes and gather the samples into a |ycbcr_comp|
* while applying a swizzle required by the spec:
*
* R, G, B should respectively map to Cr, Y, Cb
*/
for (uint32_t p = 0; p < format->n_planes; p++) {
const struct anv_format_plane *plane_format = &format->planes[p];
nir_ssa_def *plane_sample = create_plane_tex_instr_implicit(&state, p);
for (uint32_t pc = 0; pc < 4; pc++) {
enum isl_channel_select ycbcr_swizzle =
swizzle_channel(plane_format->ycbcr_swizzle, pc);
if (ycbcr_swizzle == ISL_CHANNEL_SELECT_ZERO)
continue;
unsigned ycbcr_component = channel_to_component(ycbcr_swizzle);
ycbcr_comp[ycbcr_component] = nir_channel(builder, plane_sample, pc);
/* Also compute the number of bits for each component. */
const struct isl_format_layout *isl_layout =
isl_format_get_layout(plane_format->isl_format);
ycbcr_bpcs[ycbcr_component] = isl_layout->channels_array[pc].bits;
}
}
/* Now remaps components to the order specified by the conversion. */
nir_ssa_def *swizzled_comp[4] = { NULL, };
uint32_t swizzled_bpcs[4] = { 0, };
for (uint32_t i = 0; i < ARRAY_SIZE(state.conversion->mapping); i++) {
/* Maps to components in |ycbcr_comp| */
static const uint32_t swizzle_mapping[] = {
[VK_COMPONENT_SWIZZLE_ZERO] = 4,
[VK_COMPONENT_SWIZZLE_ONE] = 3,
[VK_COMPONENT_SWIZZLE_R] = 0,
[VK_COMPONENT_SWIZZLE_G] = 1,
[VK_COMPONENT_SWIZZLE_B] = 2,
[VK_COMPONENT_SWIZZLE_A] = 3,
};
const VkComponentSwizzle m = state.conversion->mapping[i];
if (m == VK_COMPONENT_SWIZZLE_IDENTITY) {
swizzled_comp[i] = ycbcr_comp[i];
swizzled_bpcs[i] = ycbcr_bpcs[i];
} else {
swizzled_comp[i] = ycbcr_comp[swizzle_mapping[m]];
swizzled_bpcs[i] = ycbcr_bpcs[swizzle_mapping[m]];
}
}
nir_ssa_def *result = nir_vec(builder, swizzled_comp, 4);
if (state.conversion->ycbcr_model != VK_SAMPLER_YCBCR_MODEL_CONVERSION_RGB_IDENTITY)
result = convert_ycbcr(&state, result, swizzled_bpcs);
nir_ssa_def_rewrite_uses(&tex->dest.ssa, nir_src_for_ssa(result));
nir_instr_remove(&tex->instr);
return true;
}
bool
anv_nir_lower_ycbcr_textures(nir_shader *shader,
struct anv_pipeline_layout *layout)
{
bool progress = false;
nir_foreach_function(function, shader) {
if (!function->impl)
continue;
bool function_progress = false;
nir_builder builder;
nir_builder_init(&builder, function->impl);
nir_foreach_block(block, function->impl) {
nir_foreach_instr_safe(instr, block) {
if (instr->type != nir_instr_type_tex)
continue;
nir_tex_instr *tex = nir_instr_as_tex(instr);
function_progress |= try_lower_tex_ycbcr(layout, &builder, tex);
}
}
if (function_progress) {
nir_metadata_preserve(function->impl,
nir_metadata_block_index |
nir_metadata_dominance);
}
progress |= function_progress;
}
return progress;
}
|