Welcome to mirror list, hosted at ThFree Co, Russian Federation.

image.h « oneapi « device « kernel « cycles « intern - git.blender.org/blender.git - Unnamed repository; edit this file 'description' to name the repository.
summaryrefslogtreecommitdiff
blob: 892558d40bf7a38227eb90c1b3298738cec1cb98 (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
/* SPDX-License-Identifier: Apache-2.0
 * Copyright 2021-2022 Intel Corporation */

CCL_NAMESPACE_BEGIN

/* For oneAPI implementation we do manual lookup and interpolation. */
/* TODO: share implementation with ../cpu/image.h. */

template<typename T> ccl_device_forceinline T tex_fetch(const TextureInfo &info, int index)
{
  return reinterpret_cast<ccl_global T *>(info.data)[index];
}

ccl_device_inline int svm_image_texture_wrap_periodic(int x, int width)
{
  x %= width;
  if (x < 0)
    x += width;
  return x;
}

ccl_device_inline int svm_image_texture_wrap_clamp(int x, int width)
{
  return clamp(x, 0, width - 1);
}

ccl_device_inline float4 svm_image_texture_read(const TextureInfo &info, int x, int y, int z)
{
  const int data_offset = x + info.width * y + info.width * info.height * z;
  const int texture_type = info.data_type;

  /* Float4 */
  if (texture_type == IMAGE_DATA_TYPE_FLOAT4) {
    return tex_fetch<float4>(info, data_offset);
  }
  /* Byte4 */
  else if (texture_type == IMAGE_DATA_TYPE_BYTE4) {
    uchar4 r = tex_fetch<uchar4>(info, data_offset);
    float f = 1.0f / 255.0f;
    return make_float4(r.x * f, r.y * f, r.z * f, r.w * f);
  }
  /* Ushort4 */
  else if (texture_type == IMAGE_DATA_TYPE_USHORT4) {
    ushort4 r = tex_fetch<ushort4>(info, data_offset);
    float f = 1.0f / 65535.f;
    return make_float4(r.x * f, r.y * f, r.z * f, r.w * f);
  }
  /* Float */
  else if (texture_type == IMAGE_DATA_TYPE_FLOAT) {
    float f = tex_fetch<float>(info, data_offset);
    return make_float4(f, f, f, 1.0f);
  }
  /* UShort */
  else if (texture_type == IMAGE_DATA_TYPE_USHORT) {
    ushort r = tex_fetch<ushort>(info, data_offset);
    float f = r * (1.0f / 65535.0f);
    return make_float4(f, f, f, 1.0f);
  }
  else if (texture_type == IMAGE_DATA_TYPE_HALF) {
    float f = tex_fetch<half>(info, data_offset);
    return make_float4(f, f, f, 1.0f);
  }
  else if (texture_type == IMAGE_DATA_TYPE_HALF4) {
    half4 r = tex_fetch<half4>(info, data_offset);
    return make_float4(r.x, r.y, r.z, r.w);
  }
  /* Byte */
  else {
    uchar r = tex_fetch<uchar>(info, data_offset);
    float f = r * (1.0f / 255.0f);
    return make_float4(f, f, f, 1.0f);
  }
}

ccl_device_inline float4 svm_image_texture_read_2d(int id, int x, int y)
{
  const TextureInfo &info = kernel_data_fetch(texture_info, id);

  /* Wrap */
  if (info.extension == EXTENSION_REPEAT) {
    x = svm_image_texture_wrap_periodic(x, info.width);
    y = svm_image_texture_wrap_periodic(y, info.height);
  }
  else {
    x = svm_image_texture_wrap_clamp(x, info.width);
    y = svm_image_texture_wrap_clamp(y, info.height);
  }

  return svm_image_texture_read(info, x, y, 0);
}

ccl_device_inline float4 svm_image_texture_read_3d(int id, int x, int y, int z)
{
  const TextureInfo &info = kernel_data_fetch(texture_info, id);

  /* Wrap */
  if (info.extension == EXTENSION_REPEAT) {
    x = svm_image_texture_wrap_periodic(x, info.width);
    y = svm_image_texture_wrap_periodic(y, info.height);
    z = svm_image_texture_wrap_periodic(z, info.depth);
  }
  else {
    x = svm_image_texture_wrap_clamp(x, info.width);
    y = svm_image_texture_wrap_clamp(y, info.height);
    z = svm_image_texture_wrap_clamp(z, info.depth);
  }

  return svm_image_texture_read(info, x, y, z);
}

static float svm_image_texture_frac(float x, int *ix)
{
  int i = float_to_int(x) - ((x < 0.0f) ? 1 : 0);
  *ix = i;
  return x - (float)i;
}

#define SET_CUBIC_SPLINE_WEIGHTS(u, t) \
  { \
    u[0] = (((-1.0f / 6.0f) * t + 0.5f) * t - 0.5f) * t + (1.0f / 6.0f); \
    u[1] = ((0.5f * t - 1.0f) * t) * t + (2.0f / 3.0f); \
    u[2] = ((-0.5f * t + 0.5f) * t + 0.5f) * t + (1.0f / 6.0f); \
    u[3] = (1.0f / 6.0f) * t * t * t; \
  } \
  (void)0

ccl_device float4 kernel_tex_image_interp(KernelGlobals, int id, float x, float y)
{
  const TextureInfo &info = kernel_data_fetch(texture_info, id);

  if (info.extension == EXTENSION_CLIP) {
    if (x < 0.0f || y < 0.0f || x > 1.0f || y > 1.0f) {
      return make_float4(0.0f, 0.0f, 0.0f, 0.0f);
    }
  }

  if (info.interpolation == INTERPOLATION_CLOSEST) {
    /* Closest interpolation. */
    int ix, iy;
    svm_image_texture_frac(x * info.width, &ix);
    svm_image_texture_frac(y * info.height, &iy);

    return svm_image_texture_read_2d(id, ix, iy);
  }
  else if (info.interpolation == INTERPOLATION_LINEAR) {
    /* Bilinear interpolation. */
    int ix, iy;
    float tx = svm_image_texture_frac(x * info.width - 0.5f, &ix);
    float ty = svm_image_texture_frac(y * info.height - 0.5f, &iy);

    float4 r;
    r = (1.0f - ty) * (1.0f - tx) * svm_image_texture_read_2d(id, ix, iy);
    r += (1.0f - ty) * tx * svm_image_texture_read_2d(id, ix + 1, iy);
    r += ty * (1.0f - tx) * svm_image_texture_read_2d(id, ix, iy + 1);
    r += ty * tx * svm_image_texture_read_2d(id, ix + 1, iy + 1);
    return r;
  }
  else {
    /* Bicubic interpolation. */
    int ix, iy;
    float tx = svm_image_texture_frac(x * info.width - 0.5f, &ix);
    float ty = svm_image_texture_frac(y * info.height - 0.5f, &iy);

    float u[4], v[4];
    SET_CUBIC_SPLINE_WEIGHTS(u, tx);
    SET_CUBIC_SPLINE_WEIGHTS(v, ty);

    float4 r = make_float4(0.0f, 0.0f, 0.0f, 0.0f);

    for (int y = 0; y < 4; y++) {
      for (int x = 0; x < 4; x++) {
        float weight = u[x] * v[y];
        r += weight * svm_image_texture_read_2d(id, ix + x - 1, iy + y - 1);
      }
    }
    return r;
  }
}

#ifdef WITH_NANOVDB
template<typename T> struct NanoVDBInterpolator {

  typedef typename nanovdb::NanoGrid<T>::AccessorType AccessorType;

  static ccl_always_inline float4 read(float r)
  {
    return make_float4(r, r, r, 1.0f);
  }

  static ccl_always_inline float4 read(nanovdb::Vec3f r)
  {
    return make_float4(r[0], r[1], r[2], 1.0f);
  }

  static ccl_always_inline float4 interp_3d_closest(const AccessorType &acc,
                                                    float x,
                                                    float y,
                                                    float z)
  {
    const nanovdb::Vec3f xyz(x, y, z);
    return read(nanovdb::SampleFromVoxels<AccessorType, 0, false>(acc)(xyz));
  }

  static ccl_always_inline float4 interp_3d_linear(const AccessorType &acc,
                                                   float x,
                                                   float y,
                                                   float z)
  {
    const nanovdb::Vec3f xyz(x - 0.5f, y - 0.5f, z - 0.5f);
    return read(nanovdb::SampleFromVoxels<AccessorType, 1, false>(acc)(xyz));
  }

  static float4 interp_3d_cubic(const AccessorType &acc, float x, float y, float z)
  {
    int ix, iy, iz;
    int nix, niy, niz;
    int pix, piy, piz;
    int nnix, nniy, nniz;
    /* Tricubic b-spline interpolation. */
    const float tx = svm_image_texture_frac(x - 0.5f, &ix);
    const float ty = svm_image_texture_frac(y - 0.5f, &iy);
    const float tz = svm_image_texture_frac(z - 0.5f, &iz);
    pix = ix - 1;
    piy = iy - 1;
    piz = iz - 1;
    nix = ix + 1;
    niy = iy + 1;
    niz = iz + 1;
    nnix = ix + 2;
    nniy = iy + 2;
    nniz = iz + 2;

    const int xc[4] = {pix, ix, nix, nnix};
    const int yc[4] = {piy, iy, niy, nniy};
    const int zc[4] = {piz, iz, niz, nniz};
    float u[4], v[4], w[4];

    /* Some helper macro to keep code reasonable size,
     * let compiler to inline all the matrix multiplications.
     */
#  define DATA(x, y, z) (read(acc.getValue(nanovdb::Coord(xc[x], yc[y], zc[z]))))
#  define COL_TERM(col, row) \
    (v[col] * (u[0] * DATA(0, col, row) + u[1] * DATA(1, col, row) + u[2] * DATA(2, col, row) + \
               u[3] * DATA(3, col, row)))
#  define ROW_TERM(row) \
    (w[row] * (COL_TERM(0, row) + COL_TERM(1, row) + COL_TERM(2, row) + COL_TERM(3, row)))

    SET_CUBIC_SPLINE_WEIGHTS(u, tx);
    SET_CUBIC_SPLINE_WEIGHTS(v, ty);
    SET_CUBIC_SPLINE_WEIGHTS(w, tz);

    /* Actual interpolation. */
    return ROW_TERM(0) + ROW_TERM(1) + ROW_TERM(2) + ROW_TERM(3);

#  undef COL_TERM
#  undef ROW_TERM
#  undef DATA
  }

  static ccl_always_inline float4
  interp_3d(const TextureInfo &info, float x, float y, float z, int interp)
  {
    using namespace nanovdb;

    NanoGrid<T> *const grid = (NanoGrid<T> *)info.data;
    AccessorType acc = grid->getAccessor();

    switch ((interp == INTERPOLATION_NONE) ? info.interpolation : interp) {
      case INTERPOLATION_CLOSEST:
        return interp_3d_closest(acc, x, y, z);
      case INTERPOLATION_LINEAR:
        return interp_3d_linear(acc, x, y, z);
      default:
        return interp_3d_cubic(acc, x, y, z);
    }
  }
};
#endif /* WITH_NANOVDB */

ccl_device float4 kernel_tex_image_interp_3d(KernelGlobals, int id, float3 P, int interp)
{
  const TextureInfo &info = kernel_data_fetch(texture_info, id);

  if (info.use_transform_3d) {
    Transform tfm = info.transform_3d;
    P = transform_point(&tfm, P);
  }

  float x = P.x;
  float y = P.y;
  float z = P.z;

  uint interpolation = (interp == INTERPOLATION_NONE) ? info.interpolation : interp;

#ifdef WITH_NANOVDB
  if (info.data_type == IMAGE_DATA_TYPE_NANOVDB_FLOAT) {
    return NanoVDBInterpolator<float>::interp_3d(info, x, y, z, interpolation);
  }
  else if (info.data_type == IMAGE_DATA_TYPE_NANOVDB_FLOAT3) {
    return NanoVDBInterpolator<nanovdb::Vec3f>::interp_3d(info, x, y, z, interpolation);
  }
  else if (info.data_type == IMAGE_DATA_TYPE_NANOVDB_FPN) {
    return NanoVDBInterpolator<nanovdb::FpN>::interp_3d(info, x, y, z, interpolation);
  }
  else if (info.data_type == IMAGE_DATA_TYPE_NANOVDB_FP16) {
    return NanoVDBInterpolator<nanovdb::Fp16>::interp_3d(info, x, y, z, interpolation);
  }
#else
  if (info.data_type == IMAGE_DATA_TYPE_NANOVDB_FLOAT ||
      info.data_type == IMAGE_DATA_TYPE_NANOVDB_FLOAT3 ||
      info.data_type == IMAGE_DATA_TYPE_NANOVDB_FPN ||
      info.data_type == IMAGE_DATA_TYPE_NANOVDB_FP16) {
    return make_float4(
        TEX_IMAGE_MISSING_R, TEX_IMAGE_MISSING_G, TEX_IMAGE_MISSING_B, TEX_IMAGE_MISSING_A);
  }
#endif
  else {
    if (info.extension == EXTENSION_CLIP) {
      if (x < 0.0f || y < 0.0f || z < 0.0f || x > 1.0f || y > 1.0f || z > 1.0f) {
        return make_float4(0.0f, 0.0f, 0.0f, 0.0f);
      }
    }

    x *= info.width;
    y *= info.height;
    z *= info.depth;
  }

  if (interpolation == INTERPOLATION_CLOSEST) {
    /* Closest interpolation. */
    int ix, iy, iz;
    svm_image_texture_frac(x, &ix);
    svm_image_texture_frac(y, &iy);
    svm_image_texture_frac(z, &iz);

    return svm_image_texture_read_3d(id, ix, iy, iz);
  }
  else if (interpolation == INTERPOLATION_LINEAR) {
    /* Trilinear interpolation. */
    int ix, iy, iz;
    float tx = svm_image_texture_frac(x - 0.5f, &ix);
    float ty = svm_image_texture_frac(y - 0.5f, &iy);
    float tz = svm_image_texture_frac(z - 0.5f, &iz);

    float4 r;
    r = (1.0f - tz) * (1.0f - ty) * (1.0f - tx) * svm_image_texture_read_3d(id, ix, iy, iz);
    r += (1.0f - tz) * (1.0f - ty) * tx * svm_image_texture_read_3d(id, ix + 1, iy, iz);
    r += (1.0f - tz) * ty * (1.0f - tx) * svm_image_texture_read_3d(id, ix, iy + 1, iz);
    r += (1.0f - tz) * ty * tx * svm_image_texture_read_3d(id, ix + 1, iy + 1, iz);

    r += tz * (1.0f - ty) * (1.0f - tx) * svm_image_texture_read_3d(id, ix, iy, iz + 1);
    r += tz * (1.0f - ty) * tx * svm_image_texture_read_3d(id, ix + 1, iy, iz + 1);
    r += tz * ty * (1.0f - tx) * svm_image_texture_read_3d(id, ix, iy + 1, iz + 1);
    r += tz * ty * tx * svm_image_texture_read_3d(id, ix + 1, iy + 1, iz + 1);
    return r;
  }
  else {
    /* Tricubic interpolation. */
    int ix, iy, iz;
    float tx = svm_image_texture_frac(x - 0.5f, &ix);
    float ty = svm_image_texture_frac(y - 0.5f, &iy);
    float tz = svm_image_texture_frac(z - 0.5f, &iz);

    float u[4], v[4], w[4];
    SET_CUBIC_SPLINE_WEIGHTS(u, tx);
    SET_CUBIC_SPLINE_WEIGHTS(v, ty);
    SET_CUBIC_SPLINE_WEIGHTS(w, tz);

    float4 r = make_float4(0.0f, 0.0f, 0.0f, 0.0f);

    for (int z = 0; z < 4; z++) {
      for (int y = 0; y < 4; y++) {
        for (int x = 0; x < 4; x++) {
          float weight = u[x] * v[y] * w[z];
          r += weight * svm_image_texture_read_3d(id, ix + x - 1, iy + y - 1, iz + z - 1);
        }
      }
    }
    return r;
  }
}

#undef SET_CUBIC_SPLINE_WEIGHTS

CCL_NAMESPACE_END