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
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
|
/*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
CCL_NAMESPACE_BEGIN
#ifdef __HAIR__
/* curve attributes */
ccl_device float curve_attribute_float(KernelGlobals *kg, const ShaderData *sd, AttributeElement elem, int offset, float *dx, float *dy)
{
if(elem == ATTR_ELEMENT_CURVE) {
#ifdef __RAY_DIFFERENTIALS__
if(dx) *dx = 0.0f;
if(dy) *dy = 0.0f;
#endif
return kernel_tex_fetch(__attributes_float, offset + sd->prim);
}
else if(elem == ATTR_ELEMENT_CURVE_KEY) {
float4 curvedata = kernel_tex_fetch(__curves, sd->prim);
int k0 = __float_as_int(curvedata.x) + sd->segment;
int k1 = k0 + 1;
float f0 = kernel_tex_fetch(__attributes_float, offset + k0);
float f1 = kernel_tex_fetch(__attributes_float, offset + k1);
#ifdef __RAY_DIFFERENTIALS__
if(dx) *dx = sd->du.dx*(f1 - f0);
if(dy) *dy = 0.0f;
#endif
return (1.0f - sd->u)*f0 + sd->u*f1;
}
else {
#ifdef __RAY_DIFFERENTIALS__
if(dx) *dx = 0.0f;
if(dy) *dy = 0.0f;
#endif
return 0.0f;
}
}
ccl_device float3 curve_attribute_float3(KernelGlobals *kg, const ShaderData *sd, AttributeElement elem, int offset, float3 *dx, float3 *dy)
{
if(elem == ATTR_ELEMENT_CURVE) {
/* idea: we can't derive any useful differentials here, but for tiled
* mipmap image caching it would be useful to avoid reading the highest
* detail level always. maybe a derivative based on the hair density
* could be computed somehow? */
#ifdef __RAY_DIFFERENTIALS__
if(dx) *dx = make_float3(0.0f, 0.0f, 0.0f);
if(dy) *dy = make_float3(0.0f, 0.0f, 0.0f);
#endif
return float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + sd->prim));
}
else if(elem == ATTR_ELEMENT_CURVE_KEY) {
float4 curvedata = kernel_tex_fetch(__curves, sd->prim);
int k0 = __float_as_int(curvedata.x) + sd->segment;
int k1 = k0 + 1;
float3 f0 = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + k0));
float3 f1 = float4_to_float3(kernel_tex_fetch(__attributes_float3, offset + k1));
#ifdef __RAY_DIFFERENTIALS__
if(dx) *dx = sd->du.dx*(f1 - f0);
if(dy) *dy = make_float3(0.0f, 0.0f, 0.0f);
#endif
return (1.0f - sd->u)*f0 + sd->u*f1;
}
else {
#ifdef __RAY_DIFFERENTIALS__
if(dx) *dx = make_float3(0.0f, 0.0f, 0.0f);
if(dy) *dy = make_float3(0.0f, 0.0f, 0.0f);
#endif
return make_float3(0.0f, 0.0f, 0.0f);
}
}
/* hair info node functions */
ccl_device float curve_thickness(KernelGlobals *kg, ShaderData *sd)
{
float r = 0.0f;
if(sd->segment != ~0) {
float4 curvedata = kernel_tex_fetch(__curves, sd->prim);
int k0 = __float_as_int(curvedata.x) + sd->segment;
int k1 = k0 + 1;
float4 P1 = kernel_tex_fetch(__curve_keys, k0);
float4 P2 = kernel_tex_fetch(__curve_keys, k1);
r = (P2.w - P1.w) * sd->u + P1.w;
}
return r*2.0f;
}
ccl_device float3 curve_tangent_normal(KernelGlobals *kg, ShaderData *sd)
{
float3 tgN = make_float3(0.0f,0.0f,0.0f);
if(sd->segment != ~0) {
tgN = -(-sd->I - sd->dPdu * (dot(sd->dPdu,-sd->I) / len_squared(sd->dPdu)));
tgN = normalize(tgN);
/* need to find suitable scaled gd for corrected normal */
#if 0
tgN = normalize(tgN - gd * sd->dPdu);
#endif
}
return tgN;
}
#endif
#ifdef __HAIR__
ccl_device_inline void curvebounds(float *lower, float *upper, float *extremta, float *extrema, float *extremtb, float *extremb, float p0, float p1, float p2, float p3)
{
float halfdiscroot = (p2 * p2 - 3 * p3 * p1);
float ta = -1.0f;
float tb = -1.0f;
*extremta = -1.0f;
*extremtb = -1.0f;
*upper = p0;
*lower = p0 + p1 + p2 + p3;
*extrema = *upper;
*extremb = *lower;
if(*lower >= *upper) {
*upper = *lower;
*lower = p0;
}
if(halfdiscroot >= 0) {
halfdiscroot = sqrt(halfdiscroot);
ta = (-p2 - halfdiscroot) / (3 * p3);
tb = (-p2 + halfdiscroot) / (3 * p3);
}
float t2;
float t3;
if(ta > 0.0f && ta < 1.0f) {
t2 = ta * ta;
t3 = t2 * ta;
*extremta = ta;
*extrema = p3 * t3 + p2 * t2 + p1 * ta + p0;
if(*extrema > *upper) {
*upper = *extrema;
}
if(*extrema < *lower) {
*lower = *extrema;
}
}
if(tb > 0.0f && tb < 1.0f) {
t2 = tb * tb;
t3 = t2 * tb;
*extremtb = tb;
*extremb = p3 * t3 + p2 * t2 + p1 * tb + p0;
if(*extremb >= *upper) {
*upper = *extremb;
}
if(*extremb <= *lower) {
*lower = *extremb;
}
}
}
#ifdef __KERNEL_SSE2__
ccl_device_inline __m128 transform_point_T3(const __m128 t[3], const __m128 &a)
{
return fma(broadcast<0>(a), t[0], fma(broadcast<1>(a), t[1], _mm_mul_ps(broadcast<2>(a), t[2])));
}
#endif
#ifdef __KERNEL_SSE2__
/* Pass P and idir by reference to aligned vector */
ccl_device_inline bool bvh_cardinal_curve_intersect(KernelGlobals *kg, Intersection *isect,
const float3 &P, const float3 &idir, uint visibility, int object, int curveAddr, float time, int segment, uint *lcg_state, float difl, float extmax)
#else
ccl_device_inline bool bvh_cardinal_curve_intersect(KernelGlobals *kg, Intersection *isect,
float3 P, float3 idir, uint visibility, int object, int curveAddr, float time, int segment, uint *lcg_state, float difl, float extmax)
#endif
{
float epsilon = 0.0f;
float r_st, r_en;
int depth = kernel_data.curve.subdivisions;
int flags = kernel_data.curve.curveflags;
int prim = kernel_tex_fetch(__prim_index, curveAddr);
#ifdef __KERNEL_SSE2__
__m128 vdir = _mm_div_ps(_mm_set1_ps(1.0f), load_m128(idir));
__m128 vcurve_coef[4];
const float3 *curve_coef = (float3 *)vcurve_coef;
{
__m128 dtmp = _mm_mul_ps(vdir, vdir);
__m128 d_ss = _mm_sqrt_ss(_mm_add_ss(dtmp, broadcast<2>(dtmp)));
__m128 rd_ss = _mm_div_ss(_mm_set_ss(1.0f), d_ss);
__m128i v00vec = _mm_load_si128((__m128i *)&kg->__curves.data[prim]);
int2 &v00 = (int2 &)v00vec;
int k0 = v00.x + segment;
int k1 = k0 + 1;
int ka = max(k0 - 1, v00.x);
int kb = min(k1 + 1, v00.x + v00.y - 1);
__m128 P_curve[4];
P_curve[0] = _mm_load_ps(&kg->__curve_keys.data[ka].x);
P_curve[1] = _mm_load_ps(&kg->__curve_keys.data[k0].x);
P_curve[2] = _mm_load_ps(&kg->__curve_keys.data[k1].x);
P_curve[3] = _mm_load_ps(&kg->__curve_keys.data[kb].x);
__m128 rd_sgn = set_sign_bit<0, 1, 1, 1>(broadcast<0>(rd_ss));
__m128 mul_zxxy = _mm_mul_ps(shuffle<2, 0, 0, 1>(vdir), rd_sgn);
__m128 mul_yz = _mm_mul_ps(shuffle<1, 2, 1, 2>(vdir), mul_zxxy);
__m128 mul_shuf = shuffle<0, 1, 2, 3>(mul_zxxy, mul_yz);
__m128 vdir0 = _mm_and_ps(vdir, _mm_castsi128_ps(_mm_setr_epi32(0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0)));
__m128 htfm0 = shuffle<0, 2, 0, 3>(mul_shuf, vdir0);
__m128 htfm1 = shuffle<1, 0, 1, 3>(_mm_set_ss(_mm_cvtss_f32(d_ss)), vdir0);
__m128 htfm2 = shuffle<1, 3, 2, 3>(mul_shuf, vdir0);
__m128 htfm[] = { htfm0, htfm1, htfm2 };
__m128 vP = load_m128(P);
__m128 p0 = transform_point_T3(htfm, _mm_sub_ps(P_curve[0], vP));
__m128 p1 = transform_point_T3(htfm, _mm_sub_ps(P_curve[1], vP));
__m128 p2 = transform_point_T3(htfm, _mm_sub_ps(P_curve[2], vP));
__m128 p3 = transform_point_T3(htfm, _mm_sub_ps(P_curve[3], vP));
float fc = 0.71f;
__m128 vfc = _mm_set1_ps(fc);
__m128 vfcxp3 = _mm_mul_ps(vfc, p3);
vcurve_coef[0] = p1;
vcurve_coef[1] = _mm_mul_ps(vfc, _mm_sub_ps(p2, p0));
vcurve_coef[2] = fma(_mm_set1_ps(fc * 2.0f), p0, fma(_mm_set1_ps(fc - 3.0f), p1, fms(_mm_set1_ps(3.0f - 2.0f * fc), p2, vfcxp3)));
vcurve_coef[3] = fms(_mm_set1_ps(fc - 2.0f), _mm_sub_ps(p2, p1), fms(vfc, p0, vfcxp3));
r_st = ((float4 &)P_curve[1]).w;
r_en = ((float4 &)P_curve[2]).w;
}
#else
float3 curve_coef[4];
/* curve Intersection check */
float3 dir = 1.0f/idir;
/* obtain curve parameters */
{
/* ray transform created - this should be created at beginning of intersection loop */
Transform htfm;
float d = sqrtf(dir.x * dir.x + dir.z * dir.z);
htfm = make_transform(
dir.z / d, 0, -dir.x /d, 0,
-dir.x * dir.y /d, d, -dir.y * dir.z /d, 0,
dir.x, dir.y, dir.z, 0,
0, 0, 0, 1);
float4 v00 = kernel_tex_fetch(__curves, prim);
int k0 = __float_as_int(v00.x) + segment;
int k1 = k0 + 1;
int ka = max(k0 - 1,__float_as_int(v00.x));
int kb = min(k1 + 1,__float_as_int(v00.x) + __float_as_int(v00.y) - 1);
float4 P_curve[4];
P_curve[0] = kernel_tex_fetch(__curve_keys, ka);
P_curve[1] = kernel_tex_fetch(__curve_keys, k0);
P_curve[2] = kernel_tex_fetch(__curve_keys, k1);
P_curve[3] = kernel_tex_fetch(__curve_keys, kb);
float3 p0 = transform_point(&htfm, float4_to_float3(P_curve[0]) - P);
float3 p1 = transform_point(&htfm, float4_to_float3(P_curve[1]) - P);
float3 p2 = transform_point(&htfm, float4_to_float3(P_curve[2]) - P);
float3 p3 = transform_point(&htfm, float4_to_float3(P_curve[3]) - P);
float fc = 0.71f;
curve_coef[0] = p1;
curve_coef[1] = -fc*p0 + fc*p2;
curve_coef[2] = 2.0f * fc * p0 + (fc - 3.0f) * p1 + (3.0f - 2.0f * fc) * p2 - fc * p3;
curve_coef[3] = -fc * p0 + (2.0f - fc) * p1 + (fc - 2.0f) * p2 + fc * p3;
r_st = P_curve[1].w;
r_en = P_curve[2].w;
}
#endif
float r_curr = max(r_st, r_en);
if((flags & CURVE_KN_RIBBONS) || !(flags & CURVE_KN_BACKFACING))
epsilon = 2 * r_curr;
/* find bounds - this is slow for cubic curves */
float upper, lower;
float zextrem[4];
curvebounds(&lower, &upper, &zextrem[0], &zextrem[1], &zextrem[2], &zextrem[3], curve_coef[0].z, curve_coef[1].z, curve_coef[2].z, curve_coef[3].z);
if(lower - r_curr > isect->t || upper + r_curr < epsilon)
return false;
/* minimum width extension */
float mw_extension = min(difl * fabsf(upper), extmax);
float r_ext = mw_extension + r_curr;
float xextrem[4];
curvebounds(&lower, &upper, &xextrem[0], &xextrem[1], &xextrem[2], &xextrem[3], curve_coef[0].x, curve_coef[1].x, curve_coef[2].x, curve_coef[3].x);
if(lower > r_ext || upper < -r_ext)
return false;
float yextrem[4];
curvebounds(&lower, &upper, &yextrem[0], &yextrem[1], &yextrem[2], &yextrem[3], curve_coef[0].y, curve_coef[1].y, curve_coef[2].y, curve_coef[3].y);
if(lower > r_ext || upper < -r_ext)
return false;
/* setup recurrent loop */
int level = 1 << depth;
int tree = 0;
float resol = 1.0f / (float)level;
bool hit = false;
/* begin loop */
while(!(tree >> (depth))) {
float i_st = tree * resol;
float i_en = i_st + (level * resol);
#ifdef __KERNEL_SSE2__
__m128 vi_st = _mm_set1_ps(i_st), vi_en = _mm_set1_ps(i_en);
__m128 vp_st = fma(fma(fma(vcurve_coef[3], vi_st, vcurve_coef[2]), vi_st, vcurve_coef[1]), vi_st, vcurve_coef[0]);
__m128 vp_en = fma(fma(fma(vcurve_coef[3], vi_en, vcurve_coef[2]), vi_en, vcurve_coef[1]), vi_en, vcurve_coef[0]);
__m128 vbmin = _mm_min_ps(vp_st, vp_en);
__m128 vbmax = _mm_max_ps(vp_st, vp_en);
float3 &bmin = (float3 &)vbmin, &bmax = (float3 &)vbmax;
float &bminx = bmin.x, &bminy = bmin.y, &bminz = bmin.z;
float &bmaxx = bmax.x, &bmaxy = bmax.y, &bmaxz = bmax.z;
float3 &p_st = (float3 &)vp_st, &p_en = (float3 &)vp_en;
#else
float3 p_st = ((curve_coef[3] * i_st + curve_coef[2]) * i_st + curve_coef[1]) * i_st + curve_coef[0];
float3 p_en = ((curve_coef[3] * i_en + curve_coef[2]) * i_en + curve_coef[1]) * i_en + curve_coef[0];
float bminx = min(p_st.x, p_en.x);
float bmaxx = max(p_st.x, p_en.x);
float bminy = min(p_st.y, p_en.y);
float bmaxy = max(p_st.y, p_en.y);
float bminz = min(p_st.z, p_en.z);
float bmaxz = max(p_st.z, p_en.z);
#endif
if(xextrem[0] >= i_st && xextrem[0] <= i_en) {
bminx = min(bminx,xextrem[1]);
bmaxx = max(bmaxx,xextrem[1]);
}
if(xextrem[2] >= i_st && xextrem[2] <= i_en) {
bminx = min(bminx,xextrem[3]);
bmaxx = max(bmaxx,xextrem[3]);
}
if(yextrem[0] >= i_st && yextrem[0] <= i_en) {
bminy = min(bminy,yextrem[1]);
bmaxy = max(bmaxy,yextrem[1]);
}
if(yextrem[2] >= i_st && yextrem[2] <= i_en) {
bminy = min(bminy,yextrem[3]);
bmaxy = max(bmaxy,yextrem[3]);
}
if(zextrem[0] >= i_st && zextrem[0] <= i_en) {
bminz = min(bminz,zextrem[1]);
bmaxz = max(bmaxz,zextrem[1]);
}
if(zextrem[2] >= i_st && zextrem[2] <= i_en) {
bminz = min(bminz,zextrem[3]);
bmaxz = max(bmaxz,zextrem[3]);
}
float r1 = r_st + (r_en - r_st) * i_st;
float r2 = r_st + (r_en - r_st) * i_en;
r_curr = max(r1, r2);
mw_extension = min(difl * fabsf(bmaxz), extmax);
float r_ext = mw_extension + r_curr;
float coverage = 1.0f;
if (bminz - r_curr > isect->t || bmaxz + r_curr < epsilon || bminx > r_ext|| bmaxx < -r_ext|| bminy > r_ext|| bmaxy < -r_ext) {
/* the bounding box does not overlap the square centered at O */
tree += level;
level = tree & -tree;
}
else if (level == 1) {
/* the maximum recursion depth is reached.
* check if dP0.(Q-P0)>=0 and dPn.(Pn-Q)>=0.
* dP* is reversed if necessary.*/
float t = isect->t;
float u = 0.0f;
if(flags & CURVE_KN_RIBBONS) {
float3 tg = (p_en - p_st);
float w = tg.x * tg.x + tg.y * tg.y;
if (w == 0) {
tree++;
level = tree & -tree;
continue;
}
w = -(p_st.x * tg.x + p_st.y * tg.y) / w;
w = clamp((float)w, 0.0f, 1.0f);
/* compute u on the curve segment */
u = i_st * (1 - w) + i_en * w;
r_curr = r_st + (r_en - r_st) * u;
/* compare x-y distances */
float3 p_curr = ((curve_coef[3] * u + curve_coef[2]) * u + curve_coef[1]) * u + curve_coef[0];
float3 dp_st = (3 * curve_coef[3] * i_st + 2 * curve_coef[2]) * i_st + curve_coef[1];
if (dot(tg, dp_st)< 0)
dp_st *= -1;
if (dot(dp_st, -p_st) + p_curr.z * dp_st.z < 0) {
tree++;
level = tree & -tree;
continue;
}
float3 dp_en = (3 * curve_coef[3] * i_en + 2 * curve_coef[2]) * i_en + curve_coef[1];
if (dot(tg, dp_en) < 0)
dp_en *= -1;
if (dot(dp_en, p_en) - p_curr.z * dp_en.z < 0) {
tree++;
level = tree & -tree;
continue;
}
/* compute coverage */
float r_ext = r_curr;
coverage = 1.0f;
if(difl != 0.0f) {
mw_extension = min(difl * fabsf(bmaxz), extmax);
r_ext = mw_extension + r_curr;
float d = sqrtf(p_curr.x * p_curr.x + p_curr.y * p_curr.y);
float d0 = d - r_curr;
float d1 = d + r_curr;
if (d0 >= 0)
coverage = (min(d1 / mw_extension, 1.0f) - min(d0 / mw_extension, 1.0f)) * 0.5f;
else // inside
coverage = (min(d1 / mw_extension, 1.0f) + min(-d0 / mw_extension, 1.0f)) * 0.5f;
}
if (p_curr.x * p_curr.x + p_curr.y * p_curr.y >= r_ext * r_ext || p_curr.z <= epsilon || isect->t < p_curr.z) {
tree++;
level = tree & -tree;
continue;
}
t = p_curr.z;
}
else {
float l = len(p_en - p_st);
/* minimum width extension */
float or1 = r1;
float or2 = r2;
if(difl != 0.0f) {
mw_extension = min(len(p_st - P) * difl, extmax);
or1 = r1 < mw_extension ? mw_extension : r1;
mw_extension = min(len(p_en - P) * difl, extmax);
or2 = r2 < mw_extension ? mw_extension : r2;
}
/* --- */
float3 tg = (p_en - p_st) / l;
float gd = (or2 - or1) / l;
float difz = -dot(p_st,tg);
float cyla = 1.0f - (tg.z * tg.z * (1 + gd*gd));
float halfb = (-p_st.z - tg.z*(difz + gd*(difz*gd + or1)));
float tcentre = -halfb/cyla;
float zcentre = difz + (tg.z * tcentre);
float3 tdif = - p_st;
tdif.z += tcentre;
float tdifz = dot(tdif,tg);
float tb = 2*(tdif.z - tg.z*(tdifz + gd*(tdifz*gd + or1)));
float tc = dot(tdif,tdif) - tdifz * tdifz * (1 + gd*gd) - or1*or1 - 2*or1*tdifz*gd;
float td = tb*tb - 4*cyla*tc;
if (td < 0.0f) {
tree++;
level = tree & -tree;
continue;
}
float rootd = sqrtf(td);
float correction = ((-tb - rootd)/(2*cyla));
t = tcentre + correction;
float3 dp_st = (3 * curve_coef[3] * i_st + 2 * curve_coef[2]) * i_st + curve_coef[1];
if (dot(tg, dp_st)< 0)
dp_st *= -1;
float3 dp_en = (3 * curve_coef[3] * i_en + 2 * curve_coef[2]) * i_en + curve_coef[1];
if (dot(tg, dp_en) < 0)
dp_en *= -1;
if(flags & CURVE_KN_BACKFACING && (dot(dp_st, -p_st) + t * dp_st.z < 0 || dot(dp_en, p_en) - t * dp_en.z < 0 || isect->t < t || t <= 0.0f)) {
correction = ((-tb + rootd)/(2*cyla));
t = tcentre + correction;
}
if (dot(dp_st, -p_st) + t * dp_st.z < 0 || dot(dp_en, p_en) - t * dp_en.z < 0 || isect->t < t || t <= 0.0f) {
tree++;
level = tree & -tree;
continue;
}
float w = (zcentre + (tg.z * correction))/l;
w = clamp((float)w, 0.0f, 1.0f);
/* compute u on the curve segment */
u = i_st * (1 - w) + i_en * w;
r_curr = r1 + (r2 - r1) * w;
r_ext = or1 + (or2 - or1) * w;
coverage = r_curr/r_ext;
}
/* we found a new intersection */
/* stochastic fade from minimum width */
if(lcg_state && coverage != 1.0f) {
if(lcg_step_float(lcg_state) > coverage)
return hit;
}
#ifdef __VISIBILITY_FLAG__
/* visibility flag test. we do it here under the assumption
* that most triangles are culled by node flags */
if(kernel_tex_fetch(__prim_visibility, curveAddr) & visibility)
#endif
{
/* record intersection */
isect->prim = curveAddr;
isect->object = object;
isect->segment = segment;
isect->u = u;
isect->v = 0.0f;
/*isect->v = 1.0f - coverage; */
isect->t = t;
hit = true;
}
tree++;
level = tree & -tree;
}
else {
/* split the curve into two curves and process */
level = level >> 1;
}
}
return hit;
}
ccl_device_inline bool bvh_curve_intersect(KernelGlobals *kg, Intersection *isect,
float3 P, float3 idir, uint visibility, int object, int curveAddr, float time, int segment, uint *lcg_state, float difl, float extmax)
{
/* define few macros to minimize code duplication for SSE */
#ifndef __KERNEL_SSE2__
#define len3_squared(x) len_squared(x)
#define len3(x) len(x)
#define dot3(x, y) dot(x, y)
#endif
/* curve Intersection check */
int flags = kernel_data.curve.curveflags;
int prim = kernel_tex_fetch(__prim_index, curveAddr);
float4 v00 = kernel_tex_fetch(__curves, prim);
int cnum = __float_as_int(v00.x);
int k0 = cnum + segment;
int k1 = k0 + 1;
#ifndef __KERNEL_SSE2__
float4 P_curve[2];
P_curve[0]= kernel_tex_fetch(__curve_keys, k0);
P_curve[1]= kernel_tex_fetch(__curve_keys, k1);
float or1 = P_curve[0].w;
float or2 = P_curve[1].w;
float3 p1 = float4_to_float3(P_curve[0]);
float3 p2 = float4_to_float3(P_curve[1]);
/* minimum width extension */
float r1 = or1;
float r2 = or2;
float3 dif = P - p1;
float3 dif_second = P - p2;
if(difl != 0.0f) {
float pixelsize = min(len3(dif) * difl, extmax);
r1 = or1 < pixelsize ? pixelsize : or1;
pixelsize = min(len3(dif_second) * difl, extmax);
r2 = or2 < pixelsize ? pixelsize : or2;
}
/* --- */
float3 dir = 1.0f / idir;
float3 p21_diff = p2 - p1;
float3 sphere_dif1 = (dif + dif_second) * 0.5f;
float sphere_b_tmp = dot3(dir, sphere_dif1);
float3 sphere_dif2 = sphere_dif1 - sphere_b_tmp * dir;
#else
const __m128 p1 = _mm_load_ps(&kg->__curve_keys.data[k0].x);
const __m128 p2 = _mm_load_ps(&kg->__curve_keys.data[k1].x);
const __m128 or12 = shuffle<3, 3, 3, 3>(p1, p2);
__m128 r12 = or12;
const __m128 vP = load_m128(P);
const __m128 dif = _mm_sub_ps(vP, p1);
const __m128 dif_second = _mm_sub_ps(vP, p2);
if(difl != 0.0f) {
const __m128 len1_sq = len3_squared_splat(dif);
const __m128 len2_sq = len3_squared_splat(dif_second);
const __m128 len12 = _mm_sqrt_ps(shuffle<0, 0, 0, 0>(len1_sq, len2_sq));
const __m128 pixelsize12 = _mm_min_ps(_mm_mul_ps(len12, _mm_set1_ps(difl)), _mm_set1_ps(extmax));
r12 = _mm_max_ps(or12, pixelsize12);
}
float or1 = _mm_cvtss_f32(or12), or2 = _mm_cvtss_f32(broadcast<2>(or12));
float r1 = _mm_cvtss_f32(r12), r2 = _mm_cvtss_f32(broadcast<2>(r12));
const __m128 dir = _mm_div_ps(_mm_set1_ps(1.0f), load_m128(idir));
const __m128 p21_diff = _mm_sub_ps(p2, p1);
const __m128 sphere_dif1 = _mm_mul_ps(_mm_add_ps(dif, dif_second), _mm_set1_ps(0.5f));
const __m128 sphere_b_tmp = dot3_splat(dir, sphere_dif1);
const __m128 sphere_dif2 = fnma(sphere_b_tmp, dir, sphere_dif1);
#endif
float mr = max(r1, r2);
float l = len3(p21_diff);
float invl = 1.0f / l;
float sp_r = mr + 0.5f * l;
float sphere_b = dot3(dir, sphere_dif2);
float sdisc = sphere_b * sphere_b - len3_squared(sphere_dif2) + sp_r * sp_r;
if(sdisc < 0.0f)
return false;
/* obtain parameters and test midpoint distance for suitable modes */
#ifndef __KERNEL_SSE2__
float3 tg = p21_diff * invl;
#else
const __m128 tg = _mm_mul_ps(p21_diff, _mm_set1_ps(invl));
#endif
float gd = (r2 - r1) * invl;
float dirz = dot3(dir, tg);
float difz = dot3(dif, tg);
float a = 1.0f - (dirz*dirz*(1 + gd*gd));
float halfb = dot3(dir, dif) - dirz*(difz + gd*(difz*gd + r1));
float tcentre = -halfb/a;
float zcentre = difz + (dirz * tcentre);
if((tcentre > isect->t) && !(flags & CURVE_KN_ACCURATE))
return false;
if((zcentre < 0 || zcentre > l) && !(flags & CURVE_KN_ACCURATE) && !(flags & CURVE_KN_INTERSECTCORRECTION))
return false;
/* test minimum separation */
#ifndef __KERNEL_SSE2__
float3 cprod = cross(tg, dir);
float cprod2sq = len3_squared(cross(tg, dif));
#else
const __m128 cprod = cross(tg, dir);
float cprod2sq = len3_squared(cross_zxy(tg, dif));
#endif
float cprodsq = len3_squared(cprod);
float distscaled = dot3(cprod, dif);
if(cprodsq == 0)
distscaled = cprod2sq;
else
distscaled = (distscaled*distscaled)/cprodsq;
if(distscaled > mr*mr)
return false;
/* calculate true intersection */
#ifndef __KERNEL_SSE2__
float3 tdif = dif + tcentre * dir;
#else
const __m128 tdif = fma(_mm_set1_ps(tcentre), dir, dif);
#endif
float tdifz = dot3(tdif, tg);
float tdifma = tdifz*gd + r1;
float tb = 2*(dot3(dir, tdif) - dirz*(tdifz + gd*tdifma));
float tc = dot3(tdif, tdif) - tdifz*tdifz - tdifma*tdifma;
float td = tb*tb - 4*a*tc;
if (td < 0.0f)
return false;
float rootd = 0.0f;
float correction = 0.0f;
if(flags & CURVE_KN_ACCURATE) {
rootd = sqrtf(td);
correction = ((-tb - rootd)/(2*a));
}
float t = tcentre + correction;
if(t < isect->t) {
if(flags & CURVE_KN_INTERSECTCORRECTION) {
rootd = sqrtf(td);
correction = ((-tb - rootd)/(2*a));
t = tcentre + correction;
}
float z = zcentre + (dirz * correction);
bool backface = false;
if(flags & CURVE_KN_BACKFACING && (t < 0.0f || z < 0 || z > l)) {
backface = true;
correction = ((-tb + rootd)/(2*a));
t = tcentre + correction;
z = zcentre + (dirz * correction);
}
/* stochastic fade from minimum width */
float adjradius = or1 + z * (or2 - or1) * invl;
adjradius = adjradius / (r1 + z * gd);
if(lcg_state && adjradius != 1.0f) {
if(lcg_step_float(lcg_state) > adjradius)
return false;
}
/* --- */
if(t > 0.0f && t < isect->t && z >= 0 && z <= l) {
if (flags & CURVE_KN_ENCLOSEFILTER) {
float enc_ratio = 1.01f;
if((difz > -r1 * enc_ratio) && (dot3(dif_second, tg) < r2 * enc_ratio)) {
float a2 = 1.0f - (dirz*dirz*(1 + gd*gd*enc_ratio*enc_ratio));
float c2 = dot3(dif, dif) - difz * difz * (1 + gd*gd*enc_ratio*enc_ratio) - r1*r1*enc_ratio*enc_ratio - 2*r1*difz*gd*enc_ratio;
if(a2*c2 < 0.0f)
return false;
}
}
#ifdef __VISIBILITY_FLAG__
/* visibility flag test. we do it here under the assumption
* that most triangles are culled by node flags */
if(kernel_tex_fetch(__prim_visibility, curveAddr) & visibility)
#endif
{
/* record intersection */
isect->prim = curveAddr;
isect->object = object;
isect->segment = segment;
isect->u = z*invl;
isect->v = td/(4*a*a);
/*isect->v = 1.0f - adjradius;*/
isect->t = t;
if(backface)
isect->u = -isect->u;
return true;
}
}
}
return false;
#ifndef __KERNEL_SSE2__
#undef len3_squared
#undef len3
#undef dot3
#endif
}
#endif
#ifdef __HAIR__
ccl_device_inline float3 curvetangent(float t, float3 p0, float3 p1, float3 p2, float3 p3)
{
float fc = 0.71f;
float data[4];
float t2 = t * t;
data[0] = -3.0f * fc * t2 + 4.0f * fc * t - fc;
data[1] = 3.0f * (2.0f - fc) * t2 + 2.0f * (fc - 3.0f) * t;
data[2] = 3.0f * (fc - 2.0f) * t2 + 2.0f * (3.0f - 2.0f * fc) * t + fc;
data[3] = 3.0f * fc * t2 - 2.0f * fc * t;
return data[0] * p0 + data[1] * p1 + data[2] * p2 + data[3] * p3;
}
ccl_device_inline float3 curvepoint(float t, float3 p0, float3 p1, float3 p2, float3 p3)
{
float data[4];
float fc = 0.71f;
float t2 = t * t;
float t3 = t2 * t;
data[0] = -fc * t3 + 2.0f * fc * t2 - fc * t;
data[1] = (2.0f - fc) * t3 + (fc - 3.0f) * t2 + 1.0f;
data[2] = (fc - 2.0f) * t3 + (3.0f - 2.0f * fc) * t2 + fc * t;
data[3] = fc * t3 - fc * t2;
return data[0] * p0 + data[1] * p1 + data[2] * p2 + data[3] * p3;
}
ccl_device_inline float3 bvh_curve_refine(KernelGlobals *kg, ShaderData *sd, const Intersection *isect, const Ray *ray)
{
int flag = kernel_data.curve.curveflags;
float t = isect->t;
float3 P = ray->P;
float3 D = ray->D;
if(isect->object != ~0) {
#ifdef __OBJECT_MOTION__
Transform tfm = sd->ob_itfm;
#else
Transform tfm = object_fetch_transform(kg, isect->object, OBJECT_INVERSE_TRANSFORM);
#endif
P = transform_point(&tfm, P);
D = transform_direction(&tfm, D*t);
D = normalize_len(D, &t);
}
int prim = kernel_tex_fetch(__prim_index, isect->prim);
float4 v00 = kernel_tex_fetch(__curves, prim);
int k0 = __float_as_int(v00.x) + sd->segment;
int k1 = k0 + 1;
float3 tg;
if(flag & CURVE_KN_INTERPOLATE) {
int ka = max(k0 - 1,__float_as_int(v00.x));
int kb = min(k1 + 1,__float_as_int(v00.x) + __float_as_int(v00.y) - 1);
float4 P_curve[4];
P_curve[0] = kernel_tex_fetch(__curve_keys, ka);
P_curve[1] = kernel_tex_fetch(__curve_keys, k0);
P_curve[2] = kernel_tex_fetch(__curve_keys, k1);
P_curve[3] = kernel_tex_fetch(__curve_keys, kb);
float l = 1.0f;
tg = normalize_len(float4_to_float3(P_curve[2] - P_curve[1]), &l);
float r1 = P_curve[1].w;
float r2 = P_curve[2].w;
float gd = ((r2 - r1)/l);
P = P + D*t;
float3 p[4];
p[0] = float4_to_float3(P_curve[0]);
p[1] = float4_to_float3(P_curve[1]);
p[2] = float4_to_float3(P_curve[2]);
p[3] = float4_to_float3(P_curve[3]);
#ifdef __UV__
sd->u = isect->u;
sd->v = 0.0f;
#endif
tg = normalize(curvetangent(isect->u, p[0], p[1], p[2], p[3]));
if(kernel_data.curve.curveflags & CURVE_KN_RIBBONS)
sd->Ng = normalize(-(D - tg * (dot(tg, D))));
else {
float3 p_curr = curvepoint(isect->u, p[0], p[1], p[2], p[3]);
sd->Ng = normalize(P - p_curr);
sd->Ng = sd->Ng - gd * tg;
sd->Ng = normalize(sd->Ng);
}
sd->N = sd->Ng;
}
else {
float4 P_curve[2];
P_curve[0]= kernel_tex_fetch(__curve_keys, k0);
P_curve[1]= kernel_tex_fetch(__curve_keys, k1);
float l = 1.0f;
tg = normalize_len(float4_to_float3(P_curve[1] - P_curve[0]), &l);
float r1 = P_curve[0].w;
float r2 = P_curve[1].w;
float gd = ((r2 - r1)/l);
P = P + D*t;
float3 dif = P - float4_to_float3(P_curve[0]);
#ifdef __UV__
sd->u = dot(dif,tg)/l;
sd->v = 0.0f;
#endif
if (flag & CURVE_KN_TRUETANGENTGNORMAL) {
sd->Ng = -(D - tg * dot(tg, D));
sd->Ng = normalize(sd->Ng);
}
else {
sd->Ng = (dif - tg * sd->u * l) / (P_curve[0].w + sd->u * l * gd);
if (gd != 0.0f) {
sd->Ng = sd->Ng - gd * tg ;
sd->Ng = normalize(sd->Ng);
}
}
sd->N = sd->Ng;
}
#ifdef __DPDU__
/* dPdu/dPdv */
sd->dPdu = tg;
sd->dPdv = cross(tg, sd->Ng);
#endif
/*add fading parameter for minimum pixel width with transparency bsdf*/
/*sd->curve_transparency = isect->v;*/
/*sd->curve_radius = sd->u * gd * l + r1;*/
if(isect->object != ~0) {
#ifdef __OBJECT_MOTION__
Transform tfm = sd->ob_tfm;
#else
Transform tfm = object_fetch_transform(kg, isect->object, OBJECT_TRANSFORM);
#endif
P = transform_point(&tfm, P);
}
return P;
}
#endif
CCL_NAMESPACE_END
|