-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathoptixtracer_textured_spline_tube.cxx
2028 lines (1768 loc) · 74 KB
/
optixtracer_textured_spline_tube.cxx
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
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
//////
//
// Includes
//
// C++ STL
#include <fstream>
// OpenGL
#include <libs/cgv_gl/gl/gl.h>
// CGV framework
#include <cgv/render/texture.h>
// local includes
#include "optix_integration.h"
#include "curveutils.h"
// OptiX/CUDA kernels
#include "cuda/optix_interface.h"
// implemented header
#include "optixtracer_textured_spline_tube.h"
//////
//
// Local helpers
//
namespace {
template <class int_type>
inline int_type ceil_unsigned (const int_type v, const int_type base) {
return ((v+base-1)/base)*base;
}
inline float get_s0 (const float b0, const float b1) {
return b0 - (b1-b0);
}
template <class T>
inline T get_s0 (const T &b0, const T &b1) {
return b0 - (b1-b0);
}
inline float get_s2 (const float b1, const float b2) {
return b2 + (b2-b1);
}
template <class T>
inline T get_s2 (const T &b1, const T &b2) {
return b2 + (b2-b1);
}
inline float get_cr0 (const float p0, const float m0, const float p1) {
const float b1 = p0 + _1o3<float>*m0;
return p1 + 6.f*(p0 - b1);
}
template <class T>
inline T get_cr0 (const T &p0, const T &m0, const T &p1) {
const T b1 = p0 + _1o3<typename T::value_type>*m0;
return p1 + 6.f*(p0 - b1);
}
inline float get_cr3 (const float p0, const float p1, const float m1) {
const float b2 = p1 - _1o3<float>*m1;
return p0 + 6.f*(p1 - b2);
}
template <class T>
inline T get_cr3 (const T &p0, const T &p1, const T &m1) {
const T b2 = p1 - _1o3<typename T::value_type> * m1;
return p0 + 6.f*(p1 - b2);
}
inline void split_quadseg (
float *b0, float *b1, float *b2, float *g0, float *g1, float *g2,
const float p0, const float m0, const float p1, const float m1
){
*b0 = p0;
*b1 = p0 + 0.2667f*m0;
*g1 = p1 - 0.2667f*m1;
*g0 = *b2 = .5f*(*b1 + *g1);
*g2 = p1;
}
template <class T>
inline void split_quadseg (
T *b0, T *b1, T *b2, T *g0, T *g1, T *g2,
const T &p0, const T &m0, const T &p1, const T &m1
){
*b0 = p0;
*b1 = p0 + 0.2667f*m0;
*g1 = p1 - 0.2667f*m1;
*g0 = *b2 = .5f*(*b1 + *g1);
*g2 = p1;
}
template <unsigned dims>
Vec<dims, float> get_curve_minimum (
const QuadraticCurve<float, dims> &bezier, const LinearRoots3<float> &extrema
){
const Vec<dims, float> bstart = evalBezier(bezier, .0f),
bend = evalBezier(bezier, 1.f);
Vec<dims, float> pmin;
for (int i=0; i<3; i++)
{
float extremum = extrema.num[i] && isBetween01(extrema.roots[i]) ?
evalBezier(bezier.row(i), extrema.roots[i])
: _posInf<float>;
pmin[i] = std::min(bstart[i], std::min(extremum, bend[i]));
}
return pmin;
}
template <unsigned dims>
Vec<dims, float> get_curve_maximum (
const QuadraticCurve<float, dims> &bezier, const LinearRoots3<float> &extrema
){
const Vec<dims, float> bstart = evalBezier(bezier, .0f),
bend = evalBezier(bezier, 1.f);
Vec<dims, float> pmax;
for (int i=0; i<3; i++)
{
float extremum = extrema.num[i] && isBetween01(extrema.roots[i]) ?
evalBezier(bezier.row(i), extrema.roots[i])
: _negInf<float>;
pmax[i] = std::max(bstart[i], std::max(extremum, bend[i]));
}
return pmax;
}
OptixAabb get_quadr_bezier_tube_aabb (
const cgv::vec3 &b0,
const cgv::vec3 &b1,
const cgv::vec3 &b2,
const float r0, const float r1, const float r2
){
// convenience shorthand
using vec3 = cgv::vec3;
// represent all curves as control matrices for more wieldy notation from here on
const QuadraticCurve3<float> b{b0, b1, b2};
const QuadraticCurve1<float> r(r0, r1, r2);
// differentiate component min/max curves
const QuadraticCurve3<float> r33({vec3(r[0]), vec3(r[1]), vec3(r[2])});
const QuadraticCurve3<float> qminus = b - r33, qplus = b + r33;
const LinearCurve3<float> dqminus = toMonomial(deriveBezier(qminus)),
dqplus = toMonomial(deriveBezier(qplus));
// determine extrema
const LinearRoots3<float> extrema = solveLinear(dqminus);
const vec3 pmin = get_curve_minimum(qminus, extrema),
pmax = get_curve_maximum(qplus, extrema);
// done!
return {pmin.x(), pmin.y(), pmin.z(), pmax.x(), pmax.y(), pmax.z()};
}
OptixPayloadType set_optix_custom_isect_payload_semantics (unsigned *semantics_storage)
{
// Payloads 0 and 1: two constant inputs for the closet-hit shader(s) defined at raygen, encoding
// a 64-bit pointer to a struct holding camera-related transformation matrices living on the stack
// of the thread running the OptiX tracing kernel.
semantics_storage[1] = semantics_storage[0] =
OPTIX_PAYLOAD_SEMANTICS_TRACE_CALLER_WRITE | OPTIX_PAYLOAD_SEMANTICS_CH_READ;
// Payloads 2 and 3: two constant inputs for the custom intersection shader(s) defined at raygen,
// encoding a 64-bit pointer to the ray-centric coordinate system transformation living on the stack
// of the thread running the OptiX tracing kernel.
semantics_storage[3] = semantics_storage[2] =
OPTIX_PAYLOAD_SEMANTICS_TRACE_CALLER_WRITE | OPTIX_PAYLOAD_SEMANTICS_IS_READ;
// Payloads 4 to 17: encodes the tracing results reported by the closest-hit or miss shaders. This
// includes tube color at hit point (1 slot), uv surface coordinates of hit point (2 slots), id of
// intersected curve segment (1 slot), hit position, surface normal and curve tangent (3x3=9 slots)
// and depth value at hit (1 slot), occupying 14 payload slots in total, which will remain unused
// until the very end of a trace, allowing OptiX to free up a lot of registers for general use.
for (unsigned i=4; i<18; i++)
semantics_storage[i] = OPTIX_PAYLOAD_SEMANTICS_TRACE_CALLER_READ
| OPTIX_PAYLOAD_SEMANTICS_MS_WRITE | OPTIX_PAYLOAD_SEMANTICS_CH_WRITE;
// return ready-made OptiX payload descriptor
return {18, semantics_storage};
}
};
//////
//
// Class implementations
//
////
// optixtracer_textured_spline_tube
optixtracer_textured_spline_tube::~optixtracer_textured_spline_tube() {}
optixtracer_textured_spline_tube::optixtracer_textured_spline_tube(
optixtracer_textured_spline_tube &&other
)
: context(other.context), accelds_mem(other.accelds_mem), prg_hit(other.prg_hit), prg_miss(other.prg_miss),
prg_raygen(other.prg_raygen), mod_shading(other.mod_shading), mod_geom(other.mod_geom), sbt(other.sbt),
lp(other.lp)
{
// redirect launch params SBT pointer
lp.sbt = &sbt;
// invalidate moved-from instance
other.lp = {};
other.context = nullptr;
other.accelds_mem = 0;
other.prg_hit = nullptr;
other.prg_miss = nullptr;
other.prg_raygen = nullptr;
other.mod_shading = nullptr;
other.mod_geom = nullptr;
other.sbt = {};
}
optixtracer_textured_spline_tube& optixtracer_textured_spline_tube::operator= (
optixtracer_textured_spline_tube &&other
)
{
// we're going to overwrite
destroy();
// take ownership of moved-in resources
context = other.context;
accelds_mem = other.accelds_mem;
prg_hit = other.prg_hit;
prg_miss = other.prg_miss;
prg_raygen = other.prg_raygen;
mod_shading = other.mod_shading;
mod_geom = other.mod_geom;
sbt = other.sbt;
lp = other.lp;
lp.sbt = &sbt; // redirect launch params SBT pointer
// invalidate moved-from instance
other.context = nullptr;
other.accelds_mem = 0;
other.prg_hit = nullptr;
other.prg_miss = nullptr;
other.prg_raygen = nullptr;
other.mod_shading = nullptr;
other.mod_geom = nullptr;
other.sbt = {};
other.lp = {};
// done!
return *this;
}
////
// optixtracer_textured_spline_tube_russig
optixtracer_textured_spline_tube_russig::optixtracer_textured_spline_tube_russig()
{}
optixtracer_textured_spline_tube_russig::optixtracer_textured_spline_tube_russig(
optixtracer_textured_spline_tube_russig &&other
)
: optixtracer_textured_spline_tube(std::move(other))
{
// we don't have any additional data, so just delegate to base...
}
optixtracer_textured_spline_tube_russig::~optixtracer_textured_spline_tube_russig()
{
// delegate to destroy method
destroy();
}
optixtracer_textured_spline_tube_russig& optixtracer_textured_spline_tube_russig::operator= (
optixtracer_textured_spline_tube_russig &&other
)
{
// we don't have any additional data, so just delegate to base
super::operator=(std::move(other));
// done!
return *this;
}
void optixtracer_textured_spline_tube_russig::destroy (void)
{
destroy_pipeline();
destroy_accelds();
context = nullptr;
}
bool optixtracer_textured_spline_tube_russig::built (void) const
{
// we _must_ be good to go if somehow we ended up with a valid pipeline
return lp.pipeline;
}
optixtracer_textured_spline_tube_russig optixtracer_textured_spline_tube_russig::build (
OptixDeviceContext context, const traj_manager<float>::render_data *render_data
)
{
optixtracer_textured_spline_tube_russig tracer;
tracer.context = context;
if (!tracer.update_accelds(render_data))
return optixtracer_textured_spline_tube_russig();
if (!tracer.update_pipeline())
return optixtracer_textured_spline_tube_russig();
tracer.lp.sbt = &tracer.sbt;
return tracer;
}
void optixtracer_textured_spline_tube_russig::destroy_accelds (void)
{
lp.accelds = 0; // ToDo: check if we can really leave the accelds handle dangling like this (there appears to be no destroy function for it?)
CUDA_SAFE_FREE(accelds_mem);
CUDA_SAFE_FREE(lp.positions);
CUDA_SAFE_FREE(lp.radii);
}
void optixtracer_textured_spline_tube_russig::destroy_pipeline (void)
{
CUDA_SAFE_FREE(lp.params); lp.params_size = 0;
OPTIX_SAFE_DESTROY_PIPELINE(lp.pipeline);
OPTIX_SAFE_DESTROY_MODULE(mod_shading);
OPTIX_SAFE_DESTROY_MODULE(mod_geom);
OPTIX_SAFE_DESTROY_PROGGROUP(prg_hit);
OPTIX_SAFE_DESTROY_PROGGROUP(prg_miss);
OPTIX_SAFE_DESTROY_PROGGROUP(prg_raygen);
sbt = {}; lp.sbt = nullptr;
}
bool optixtracer_textured_spline_tube_russig::update_accelds (const traj_manager<float>::render_data *render_data)
{
// make sure we start with a blank slate
destroy_accelds();
// use default options for simplicity
OptixAccelBuildOptions accel_options = {};
accel_options.buildFlags = OPTIX_BUILD_FLAG_ALLOW_COMPACTION;
accel_options.operation = OPTIX_BUILD_OPERATION_BUILD;
// stage geometry for upload - we don't use the available render data directly since we need to store the
// quadratic segments explicitly if we don't want to double the workload on the intersection shader
unsigned num;
std::vector<float3> positions;
std::vector<float> radii;
std::vector<OptixAabb> aabbs;
// prepare CPU-side vertex staging area
/* local scope */ {
const auto &rd_pos = render_data->positions;
const auto &rd_tan = render_data->tangents;
const auto &rd_rad = render_data->radii;
const auto &rd_idx = render_data->indices;
aabbs.reserve(rd_idx.size());
num = (unsigned)rd_idx.size()*3;
positions.reserve(num);
radii.reserve(num);
// convert data representation:
// - split Hermite curves into two quadratic Beziers
// - build axis aligned boundinx boxes (ToDo: build oriented BBs instead and use one BLAS per segment)
for (const auto &ds : render_data->datasets) for (const auto &traj : ds.trajs)
{
const unsigned num_segs = traj.n / 2;
for (unsigned i=0; i<num_segs; i++)
{
// obtain shortcuts to node data
const unsigned idx = rd_idx[traj.i0]+i;
const vec4 &t0 = rd_tan[idx], &t1 = rd_tan[idx+1];
const vec3 &p0 = rd_pos[idx], &p1 = rd_pos[idx+1],
m0 = vec3(t0), m1 = vec3(t1);
const float &r0 = rd_rad[idx], &r1 = rd_rad[idx+1];
// split hermite segment into 2 quadratic Beziers
vec3 b0, b1, b2, g0, g1, g2;
float br0, br1, br2, gr0, gr1, gr2;
split_quadseg(&b0, &b1, &b2, &g0, &g1, &g2, p0, m0, p1, m1);
split_quadseg(&br0, &br1, &br2, &gr0, &gr1, &gr2, r0, t0.w(), r1, t1.w());
// commit to staging buffers
// - 1st segment
aabbs.emplace_back(get_quadr_bezier_tube_aabb(b0, b1, b2, br0, br1, br2));
positions.emplace_back(to_float3(b0));
radii.emplace_back(r0);
positions.emplace_back(to_float3(b1));
radii.emplace_back(br1);
positions.emplace_back(to_float3(b2));
radii.emplace_back(br2);
// - 2nd segment
aabbs.emplace_back(get_quadr_bezier_tube_aabb(g0, g1, g2, gr0, gr1, gr2));
positions.emplace_back(to_float3(g0));
radii.emplace_back(gr0);
positions.emplace_back(to_float3(g1));
radii.emplace_back(gr1);
positions.emplace_back(to_float3(g2));
radii.emplace_back(gr2);
}
}
}
// track success - we don't immediately fail and return since the code in this function is not robust to failure
// (i.e. doesn't use RAII) so doing that would result in both host and device memory leaks
bool success = true;
// prepare geometry device memory
// - prelude
CUdeviceptr aabbs_dev=0;
const size_t positions_size = num*sizeof(float3), radii_size = num*sizeof(float), aabbs_size = aabbs.size()*sizeof(OptixAabb);
// - nodes
CUDA_CHECK_SET(cudaMalloc(reinterpret_cast<void**>(&lp.positions), positions_size), success);
CUDA_CHECK_SET(cudaMemcpy(reinterpret_cast<void*>(lp.positions), positions.data(), positions_size, cudaMemcpyHostToDevice), success);
CUDA_CHECK_SET(cudaMalloc(reinterpret_cast<void**>(&lp.radii), radii_size), success);
CUDA_CHECK_SET(cudaMemcpy(reinterpret_cast<void*>(lp.radii), radii.data(), radii_size, cudaMemcpyHostToDevice), success);
// - AABBs
CUDA_CHECK_SET(cudaMalloc(reinterpret_cast<void**>(&aabbs_dev), aabbs_size), success);
CUDA_CHECK_SET(cudaMemcpy(reinterpret_cast<void*>(aabbs_dev), aabbs.data(), aabbs_size, cudaMemcpyHostToDevice), success);
// OptiX accel-ds build input descriptor
// - SBT indices
const unsigned sbtindex[] = {0}, inputflags[] = {OPTIX_GEOMETRY_FLAG_DISABLE_ANYHIT}; // we only use one SBT record
CUdeviceptr sbtindex_dev = 0;
CUDA_CHECK_SET(cudaMalloc(reinterpret_cast<void**>(&sbtindex_dev), sizeof(sbtindex)), success);
CUDA_CHECK_SET(cudaMemcpy(reinterpret_cast<void*>(sbtindex_dev), sbtindex, sizeof(sbtindex), cudaMemcpyHostToDevice), success);
// - input descriptor
OptixBuildInput input_desc = {};
input_desc.type = OPTIX_BUILD_INPUT_TYPE_CUSTOM_PRIMITIVES;
input_desc.customPrimitiveArray.aabbBuffers = &aabbs_dev;
input_desc.customPrimitiveArray.flags = inputflags;
input_desc.customPrimitiveArray.numSbtRecords = 1;
input_desc.customPrimitiveArray.numPrimitives = (unsigned)aabbs.size();
input_desc.customPrimitiveArray.sbtIndexOffsetBuffer = sbtindex_dev;
input_desc.customPrimitiveArray.sbtIndexOffsetSizeInBytes = sizeof(unsigned);
// - working memory
CUdeviceptr tmpbuf = 0;
OptixAccelBufferSizes accelds_buffer_sizes = {0};
OPTIX_CHECK_SET(
optixAccelComputeMemoryUsage(context, &accel_options, &input_desc, 1/* num build inputs */, &accelds_buffer_sizes),
success
);
CUDA_CHECK_SET(cudaMalloc(reinterpret_cast<void**>(&tmpbuf), accelds_buffer_sizes.tempSizeInBytes), success);
// - output memory
CUdeviceptr mem_precompact = 0;
size_t compactsize_feedback_offset = ceil_unsigned<size_t>(accelds_buffer_sizes.outputSizeInBytes, 8);
/* |-- squeeze in a little 8-byte-aligned space for our compacted size feedback at the end of the output buffer */
CUDA_CHECK_SET(cudaMalloc(reinterpret_cast<void**>(&mem_precompact), compactsize_feedback_offset+8), success);
// - build accel-ds
OptixAccelEmitDesc emitted_prop;
emitted_prop.type = OPTIX_PROPERTY_TYPE_COMPACTED_SIZE;
emitted_prop.result = (CUdeviceptr)((char*)mem_precompact + compactsize_feedback_offset);
OPTIX_CHECK_SET(
optixAccelBuild(
context, 0/*CUDA stream*/, &accel_options, &input_desc, 1/*num build inputs*/,
tmpbuf, accelds_buffer_sizes.tempSizeInBytes, mem_precompact, accelds_buffer_sizes.outputSizeInBytes,
&lp.accelds, // <-- our acceleration datastructure!!!
&emitted_prop, 1/*num emitted properties*/
),
success
);
// - fetch our emitted property (the size we can compact to)
size_t compacted_size = accelds_buffer_sizes.outputSizeInBytes;
CUDA_CHECK_SET(cudaMemcpy(&compacted_size, (void*)emitted_prop.result, sizeof(size_t), cudaMemcpyDeviceToHost), success);
// - free scratch memory
CUDA_SAFE_FREE(tmpbuf);
// perform compaction if possible
if (compacted_size < accelds_buffer_sizes.outputSizeInBytes) {
CUDA_CHECK_SET(cudaMalloc(reinterpret_cast<void**>(&accelds_mem), compacted_size), success);
OPTIX_CHECK_SET(
optixAccelCompact(context, 0/*CUDA stream*/, lp.accelds, accelds_mem, compacted_size, &lp.accelds),
success
);
CUDA_SAFE_FREE(mem_precompact);
}
else
accelds_mem = mem_precompact;
// We can now free the aabb buffer used during build /** and the vertex
// inputs, since they are not needed by our trivial shading method **/
// (we won't consider cudaFree failing a failure of the whole function)
CUDA_SAFE_FREE(aabbs_dev);
// done!
return success;
}
bool optixtracer_textured_spline_tube_russig::update_pipeline (void)
{
////
// Prelude
// make sure we start with a blank slate
destroy_pipeline();
// CUDA/OptiX log storage
std::string compiler_log(8192, 0);
char* log = compiler_log.data();
size_t sizeof_log = compiler_log.size();
// pipeline build options
constexpr unsigned max_trace_depth = 1;
OptixPipelineCompileOptions pipeline_options = {};
// track success - we don't immediately fail and return since the code in this function is not robust to failure
// (i.e. doesn't use RAII) so doing that would result in both host and device memory leaks
bool success = true;
////
// Create modules
// fine-grained payload usage (enables OptiX to optimize register consumption)
unsigned payloads[18];
OptixPayloadType payloadType = set_optix_custom_isect_payload_semantics(payloads);
/* local scope */ {
OptixModuleCompileOptions mod_options = {};
mod_options.maxRegisterCount = OPTIX_COMPILE_DEFAULT_MAX_REGISTER_COUNT;
mod_options.optLevel = OPTIX_COMPILE_OPTIMIZATION_DEFAULT;
mod_options.debugLevel = OPTIX_COMPILE_DEBUG_LEVEL_MINIMAL;
mod_options.numPayloadTypes = 1;
mod_options.payloadTypes = &payloadType;
pipeline_options.traversableGraphFlags = OPTIX_TRAVERSABLE_GRAPH_FLAG_ALLOW_SINGLE_GAS;
pipeline_options.numPayloadValues = 0; // we use module-defined payloads
// - we report the curve parameter (1 attrib) as well as the first two bezier nodes (3 attribs each = 6). Unfortunately, the
// closest-hit shader will have to fetch the third node from global memory, as we are out of attribute registers at this point.
pipeline_options.numAttributeValues = 7;
#ifdef _DEBUG // Enables debug exceptions during optix launches. This may incur significant performance cost and should only be done during development.
pipeline_options.exceptionFlags =
OPTIX_EXCEPTION_FLAG_DEBUG | OPTIX_EXCEPTION_FLAG_TRACE_DEPTH | OPTIX_EXCEPTION_FLAG_STACK_OVERFLOW;
#else
pipeline_options.exceptionFlags = OPTIX_EXCEPTION_FLAG_NONE;
#endif
pipeline_options.pipelineLaunchParamsVariableName = "params";
pipeline_options.usesPrimitiveTypeFlags = OPTIX_PRIMITIVE_TYPE_FLAGS_CUSTOM;
const auto ptx = compile_cu2ptx("cuda/optix_curves_russig.cu", "rayc_russig", {CUDA_NVRTC_OPTIONS}, &compiler_log);
const size_t ptx_size = ptx.size();
if (!ptx_size)
{
std::cerr << "ERROR compiling OptiX device code! Log:" << std::endl
<< compiler_log << std::endl<<std::endl;
return false; // no resources allocated yet, so we can just return
}
OPTIX_CHECK_LOG_SET(
optixModuleCreateFromPTX(
context, &mod_options, &pipeline_options, ptx.data(), ptx_size, log, &sizeof_log, &mod_shading
),
log, sizeof_log, success
);
mod_geom = mod_shading; // all in one file for now
}
////
// Create program groups
/* local scope */ {
// common options
OptixProgramGroupOptions prg_options = {&payloadType};
// raygen shader
OptixProgramGroupDesc prg_raygen_desc = {};
prg_raygen_desc.kind = OPTIX_PROGRAM_GROUP_KIND_RAYGEN;
prg_raygen_desc.raygen.module = mod_shading;
prg_raygen_desc.raygen.entryFunctionName = "__raygen__basic";
OPTIX_CHECK_LOG_SET(
optixProgramGroupCreate(
context, &prg_raygen_desc, 1/*num program groups*/, &prg_options, log, &sizeof_log, &prg_raygen
),
log, sizeof_log, success
);
// miss shader
OptixProgramGroupDesc prg_miss_desc = {};
prg_miss_desc.kind = OPTIX_PROGRAM_GROUP_KIND_MISS;
prg_miss_desc.miss.module = mod_shading;
prg_miss_desc.miss.entryFunctionName = "__miss__ms";
OPTIX_CHECK_LOG_SET(
optixProgramGroupCreate(
context, &prg_miss_desc, 1/*num program groups*/, &prg_options, log, &sizeof_log, &prg_miss
),
log, sizeof_log, success
);
// hit shader group
OptixProgramGroupDesc prg_hit_desc = {};
prg_hit_desc.kind = OPTIX_PROGRAM_GROUP_KIND_HITGROUP;
prg_hit_desc.hitgroup.moduleCH = mod_shading;
prg_hit_desc.hitgroup.entryFunctionNameCH = "__closesthit__ch";
prg_hit_desc.hitgroup.moduleIS = mod_geom;
prg_hit_desc.hitgroup.entryFunctionNameIS = "__intersection__russig";
OPTIX_CHECK_LOG_SET(
optixProgramGroupCreate(
context, &prg_hit_desc, 1/*num program groups*/, &prg_options, log, &sizeof_log, &prg_hit
),
log, sizeof_log, success
);
}
////
// Link pipeline
/* local scope */ {
constexpr unsigned num_prgs = 3;
OptixProgramGroup prgs[num_prgs] = {prg_raygen, prg_miss, prg_hit};
OptixPipelineLinkOptions pipeline_linkoptions = {};
pipeline_linkoptions.maxTraceDepth = max_trace_depth;
#ifdef _DEBUG // Enables debug exceptions during optix launches. This may incur significant performance cost and should only be done during development.
pipeline_linkoptions.debugLevel = OPTIX_COMPILE_DEBUG_LEVEL_FULL;
#else
pipeline_linkoptions.debugLevel = OPTIX_COMPILE_DEBUG_LEVEL_NONE;
#endif
OPTIX_CHECK_LOG_SET(
optixPipelineCreate(
context, &pipeline_options, &pipeline_linkoptions, prgs, num_prgs, log, &sizeof_log, &lp.pipeline
),
log, sizeof_log, success
);
OptixStackSizes stacksizes = {};
for (const auto &prg : prgs)
OPTIX_CHECK_SET(optixUtilAccumulateStackSizes(prg, &stacksizes), success);
// ToDo: ??? investigate what exactly this is, why it is needed, what to best do here, etc.
uint32_t direct_callable_stack_size_from_traversal;
uint32_t direct_callable_stack_size_from_state;
uint32_t continuation_stack_size;
OPTIX_CHECK_SET(
optixUtilComputeStackSizes(
&stacksizes, max_trace_depth, 0/* maxCCDepth */, 0/* maxDCDepth */,
&direct_callable_stack_size_from_traversal,
&direct_callable_stack_size_from_state, &continuation_stack_size
),
success
);
OPTIX_CHECK_SET(
optixPipelineSetStackSize(
lp.pipeline, direct_callable_stack_size_from_traversal,
direct_callable_stack_size_from_state, continuation_stack_size,
1 // <-- max traversable depth - we only use the top-level GAS currently
),
success
);
}
////
// Set up shader binding table
/* local scope */ {
// our SBT record types
typedef sbt_record<data_raygen> sbt_record_raygen;
typedef sbt_record<data_miss> sbt_record_miss;
typedef sbt_record<data_hit> sbt_record_hit;
// prepare entries
// - raygen shaders
CUdeviceptr raygen_record;
const size_t raygen_record_size = sizeof(sbt_record_raygen);
CUDA_CHECK_SET(cudaMalloc(reinterpret_cast<void**>(&raygen_record), raygen_record_size), success);
sbt_record_raygen rg_sbt;
OPTIX_CHECK_SET(optixSbtRecordPackHeader(prg_raygen, &rg_sbt), success);
CUDA_CHECK_SET(cudaMemcpy(reinterpret_cast<void*>(raygen_record), &rg_sbt, raygen_record_size, cudaMemcpyHostToDevice), success);
// - miss shaders
CUdeviceptr miss_record;
size_t miss_record_size = sizeof(sbt_record_miss);
CUDA_CHECK_SET(cudaMalloc(reinterpret_cast<void**>(&miss_record), miss_record_size), success);
sbt_record_miss ms_sbt;
ms_sbt.data = {0.0f, 0.0f, 0.0f, 0.0f}; // background color (fully transparent black)
OPTIX_CHECK_SET(optixSbtRecordPackHeader(prg_miss, &ms_sbt), success);
CUDA_CHECK_SET(cudaMemcpy(reinterpret_cast<void*>(miss_record), &ms_sbt, miss_record_size, cudaMemcpyHostToDevice), success);
// - hit shaders
CUdeviceptr hitgroup_record;
size_t hitgroup_record_size = sizeof(sbt_record_hit);
CUDA_CHECK_SET (cudaMalloc(reinterpret_cast<void**>(&hitgroup_record), hitgroup_record_size), success);
sbt_record_hit hg_sbt;
OPTIX_CHECK_SET(optixSbtRecordPackHeader(prg_hit, &hg_sbt), success);
CUDA_CHECK_SET(cudaMemcpy(reinterpret_cast<void*>(hitgroup_record), &hg_sbt, hitgroup_record_size, cudaMemcpyHostToDevice), success);
// build up the SBT
sbt = {};
sbt.raygenRecord = raygen_record;
sbt.missRecordBase = miss_record;
sbt.missRecordStrideInBytes = sizeof(sbt_record_miss);
sbt.missRecordCount = 1;
sbt.hitgroupRecordBase = hitgroup_record;
sbt.hitgroupRecordStrideInBytes = sizeof(sbt_record_hit);
sbt.hitgroupRecordCount = 1;
}
// Create the device memory for our launch params
/* local scope */ {
bool params_malloc_success = true;
CUDA_CHECK_SET(cudaMalloc(reinterpret_cast<void**>(&lp.params), sizeof(curve_rt_params)), params_malloc_success);
if (params_malloc_success)
lp.params_size = sizeof(curve_rt_params);
else
success = false;
}
// done!
return success;
}
////
// optixtracer_textured_spline_tube_phantom
optixtracer_textured_spline_tube_phantom::optixtracer_textured_spline_tube_phantom()
{}
optixtracer_textured_spline_tube_phantom::optixtracer_textured_spline_tube_phantom(
optixtracer_textured_spline_tube_phantom &&other
)
: optixtracer_textured_spline_tube(std::move(other))
{
// we don't have any additional data, so just delegate to base...
}
optixtracer_textured_spline_tube_phantom::~optixtracer_textured_spline_tube_phantom()
{
// delegate to destroy method
destroy();
}
optixtracer_textured_spline_tube_phantom& optixtracer_textured_spline_tube_phantom::operator= (
optixtracer_textured_spline_tube_phantom &&other
)
{
// we don't have any additional data, so just delegate to base
super::operator=(std::move(other));
// done!
return *this;
}
void optixtracer_textured_spline_tube_phantom::destroy (void)
{
destroy_pipeline();
destroy_accelds();
context = nullptr;
}
bool optixtracer_textured_spline_tube_phantom::built (void) const
{
// we _must_ be good to go if somehow we ended up with a valid pipeline
return lp.pipeline;
}
optixtracer_textured_spline_tube_phantom optixtracer_textured_spline_tube_phantom::build (
OptixDeviceContext context, const traj_manager<float>::render_data *render_data
)
{
optixtracer_textured_spline_tube_phantom tracer;
tracer.context = context;
if (!tracer.update_accelds(render_data))
return optixtracer_textured_spline_tube_phantom();
if (!tracer.update_pipeline())
return optixtracer_textured_spline_tube_phantom();
tracer.lp.sbt = &tracer.sbt;
return tracer;
}
void optixtracer_textured_spline_tube_phantom::destroy_accelds (void)
{
lp.accelds = 0; // ToDo: check if we can really leave the accelds handle dangling like this (there appears to be no destroy function for it?)
CUDA_SAFE_FREE(accelds_mem);
CUDA_SAFE_FREE(lp.positions);
CUDA_SAFE_FREE(lp.radii);
}
void optixtracer_textured_spline_tube_phantom::destroy_pipeline (void)
{
CUDA_SAFE_FREE(lp.params); lp.params_size = 0;
OPTIX_SAFE_DESTROY_PIPELINE(lp.pipeline);
OPTIX_SAFE_DESTROY_MODULE(mod_shading);
OPTIX_SAFE_DESTROY_MODULE(mod_geom);
OPTIX_SAFE_DESTROY_PROGGROUP(prg_hit);
OPTIX_SAFE_DESTROY_PROGGROUP(prg_miss);
OPTIX_SAFE_DESTROY_PROGGROUP(prg_raygen);
sbt = {}; lp.sbt = nullptr;
}
bool optixtracer_textured_spline_tube_phantom::update_accelds (const traj_manager<float>::render_data *render_data)
{
// make sure we start with a blank slate
destroy_accelds();
// use default options for simplicity
OptixAccelBuildOptions accel_options = {};
accel_options.buildFlags = OPTIX_BUILD_FLAG_ALLOW_COMPACTION;
accel_options.operation = OPTIX_BUILD_OPERATION_BUILD;
// stage geometry for upload - we don't use the available render data directly since we need to store the
// quadratic segments explicitly if we don't want to double the workload on the intersection shader
unsigned num;
std::vector<float3> positions;
std::vector<float> radii;
std::vector<OptixAabb> aabbs;
// prepare CPU-side vertex staging area
/* local scope */ {
const auto &rd_pos = render_data->positions;
const auto &rd_tan = render_data->tangents;
const auto &rd_rad = render_data->radii;
const auto &rd_idx = render_data->indices;
aabbs.reserve(rd_idx.size());
num = (unsigned)rd_idx.size()*3;
positions.reserve(num);
radii.reserve(num);
// convert data representation:
// - split Hermite curves into two quadratic Beziers
// - build axis aligned boundinx boxes (ToDo: build oriented BBs instead and use one BLAS per segment)
for (const auto &ds : render_data->datasets) for (const auto &traj : ds.trajs)
{
const unsigned num_segs = traj.n / 2;
for (unsigned i=0; i<num_segs; i++)
{
// obtain shortcuts to node data
const unsigned idx = rd_idx[traj.i0]+i;
const vec4 &t0 = rd_tan[idx], &t1 = rd_tan[idx+1];
const vec3 &p0 = rd_pos[idx], &p1 = rd_pos[idx+1],
m0 = vec3(t0), m1 = vec3(t1);
const float &r0 = rd_rad[idx], &r1 = rd_rad[idx+1];
// split hermite segment into 2 quadratic Beziers
vec3 b0, b1, b2, g0, g1, g2;
float br0, br1, br2, gr0, gr1, gr2;
split_quadseg(&b0, &b1, &b2, &g0, &g1, &g2, p0, m0, p1, m1);
split_quadseg(&br0, &br1, &br2, &gr0, &gr1, &gr2, r0, t0.w(), r1, t1.w());
// commit to staging buffers
// ToDo: AABBs are optimized for swept-sphere, can be marginally tighter in case of swept-disc
// - 1st segment
aabbs.emplace_back(get_quadr_bezier_tube_aabb(b0, b1, b2, br0, br1, br2));
positions.emplace_back(to_float3(b0));
radii.emplace_back(r0);
positions.emplace_back(to_float3(b1));
radii.emplace_back(br1);
positions.emplace_back(to_float3(b2));
radii.emplace_back(br2);
// - 2nd segment
aabbs.emplace_back(get_quadr_bezier_tube_aabb(g0, g1, g2, gr0, gr1, gr2));
positions.emplace_back(to_float3(g0));
radii.emplace_back(gr0);
positions.emplace_back(to_float3(g1));
radii.emplace_back(gr1);
positions.emplace_back(to_float3(g2));
radii.emplace_back(gr2);
}
}
}
// track success - we don't immediately fail and return since the code in this function is not robust to failure
// (i.e. doesn't use RAII) so doing that would result in both host and device memory leaks
bool success = true;
// prepare geometry device memory
// - prelude
CUdeviceptr aabbs_dev=0;
const size_t positions_size = num*sizeof(float3), radii_size = num*sizeof(float), aabbs_size = aabbs.size()*sizeof(OptixAabb);
// - nodes
CUDA_CHECK_SET(cudaMalloc(reinterpret_cast<void**>(&lp.positions), positions_size), success);
CUDA_CHECK_SET(cudaMemcpy(reinterpret_cast<void*>(lp.positions), positions.data(), positions_size, cudaMemcpyHostToDevice), success);
CUDA_CHECK_SET(cudaMalloc(reinterpret_cast<void**>(&lp.radii), radii_size), success);
CUDA_CHECK_SET(cudaMemcpy(reinterpret_cast<void*>(lp.radii), radii.data(), radii_size, cudaMemcpyHostToDevice), success);
// - AABBs
CUDA_CHECK_SET(cudaMalloc(reinterpret_cast<void**>(&aabbs_dev), aabbs_size), success);
CUDA_CHECK_SET(cudaMemcpy(reinterpret_cast<void*>(aabbs_dev), aabbs.data(), aabbs_size, cudaMemcpyHostToDevice), success);
// OptiX accel-ds build input descriptor
// - SBT indices
const unsigned sbtindex[] = {0}, inputflags[] = {OPTIX_GEOMETRY_FLAG_DISABLE_ANYHIT}; // we only use one SBT record
CUdeviceptr sbtindex_dev = 0;
CUDA_CHECK_SET(cudaMalloc(reinterpret_cast<void**>(&sbtindex_dev), sizeof(sbtindex)), success);
CUDA_CHECK_SET(cudaMemcpy(reinterpret_cast<void*>(sbtindex_dev), sbtindex, sizeof(sbtindex), cudaMemcpyHostToDevice), success);
// - input descriptor
OptixBuildInput input_desc = {};
input_desc.type = OPTIX_BUILD_INPUT_TYPE_CUSTOM_PRIMITIVES;
input_desc.customPrimitiveArray.aabbBuffers = &aabbs_dev;
input_desc.customPrimitiveArray.flags = inputflags;
input_desc.customPrimitiveArray.numSbtRecords = 1;
input_desc.customPrimitiveArray.numPrimitives = (unsigned)aabbs.size();
input_desc.customPrimitiveArray.sbtIndexOffsetBuffer = sbtindex_dev;
input_desc.customPrimitiveArray.sbtIndexOffsetSizeInBytes = sizeof(unsigned);
// - working memory
CUdeviceptr tmpbuf = 0;
OptixAccelBufferSizes accelds_buffer_sizes = {0};
OPTIX_CHECK_SET(
optixAccelComputeMemoryUsage(context, &accel_options, &input_desc, 1/* num build inputs */, &accelds_buffer_sizes),
success
);
CUDA_CHECK_SET(cudaMalloc(reinterpret_cast<void**>(&tmpbuf), accelds_buffer_sizes.tempSizeInBytes), success);
// - output memory
CUdeviceptr mem_precompact = 0;
size_t compactsize_feedback_offset = ceil_unsigned<size_t>(accelds_buffer_sizes.outputSizeInBytes, 8);
/* |-- squeeze in a little 8-byte-aligned space for our compacted size feedback at the end of the output buffer */
CUDA_CHECK_SET(cudaMalloc(reinterpret_cast<void**>(&mem_precompact), compactsize_feedback_offset+8), success);
// - build accel-ds
OptixAccelEmitDesc emitted_prop;
emitted_prop.type = OPTIX_PROPERTY_TYPE_COMPACTED_SIZE;
emitted_prop.result = (CUdeviceptr)((char*)mem_precompact + compactsize_feedback_offset);
OPTIX_CHECK_SET(
optixAccelBuild(
context, 0/*CUDA stream*/, &accel_options, &input_desc, 1/*num build inputs*/,
tmpbuf, accelds_buffer_sizes.tempSizeInBytes, mem_precompact, accelds_buffer_sizes.outputSizeInBytes,
&lp.accelds, // <-- our acceleration datastructure!!!
&emitted_prop, 1/*num emitted properties*/
),
success
);
// - fetch our emitted property (the size we can compact to)
size_t compacted_size = accelds_buffer_sizes.outputSizeInBytes;
CUDA_CHECK_SET(cudaMemcpy(&compacted_size, (void*)emitted_prop.result, sizeof(size_t), cudaMemcpyDeviceToHost), success);
// - free scratch memory
CUDA_SAFE_FREE(tmpbuf);
// perform compaction if possible
if (compacted_size < accelds_buffer_sizes.outputSizeInBytes) {
CUDA_CHECK_SET(cudaMalloc(reinterpret_cast<void**>(&accelds_mem), compacted_size), success);
OPTIX_CHECK_SET(
optixAccelCompact(context, 0/*CUDA stream*/, lp.accelds, accelds_mem, compacted_size, &lp.accelds),
success
);
CUDA_SAFE_FREE(mem_precompact);
}
else
accelds_mem = mem_precompact;
// We can now free the aabb buffer used during build /** and the vertex
// inputs, since they are not needed by our trivial shading method **/
// (we won't consider cudaFree failing a failure of the whole function)
CUDA_SAFE_FREE(aabbs_dev);
// done!
return success;
}
bool optixtracer_textured_spline_tube_phantom::update_pipeline (void)
{
////
// Prelude
// make sure we start with a blank slate
destroy_pipeline();
// CUDA/OptiX log storage
std::string compiler_log(8192, 0);
char* log = compiler_log.data();
size_t sizeof_log = compiler_log.size();
// pipeline build options
constexpr unsigned max_trace_depth = 1;
OptixPipelineCompileOptions pipeline_options = {};
// track success - we don't immediately fail and return since the code in this function is not robust to failure
// (i.e. doesn't use RAII) so doing that would result in both host and device memory leaks
bool success = true;
////
// Create modules
// fine-grained payload usage (enables OptiX to optimize register consumption)
unsigned payloads[18];
OptixPayloadType payloadType = set_optix_custom_isect_payload_semantics(payloads);
/* local scope */ {
OptixModuleCompileOptions mod_options = {};
mod_options.maxRegisterCount = OPTIX_COMPILE_DEFAULT_MAX_REGISTER_COUNT;
mod_options.optLevel = OPTIX_COMPILE_OPTIMIZATION_DEFAULT;
mod_options.debugLevel = OPTIX_COMPILE_DEBUG_LEVEL_MINIMAL;
mod_options.numPayloadTypes = 1;
mod_options.payloadTypes = &payloadType;
pipeline_options.traversableGraphFlags = OPTIX_TRAVERSABLE_GRAPH_FLAG_ALLOW_SINGLE_GAS;
pipeline_options.numPayloadValues = 0; // we use module-defined payloads
// - we report the curve parameter (1 attrib) as well as the first two bezier nodes (3 attribs each = 6). Unfortunately, the
// closest-hit shader will have to fetch the third node from global memory, as we are out of attribute registers at this point.
pipeline_options.numAttributeValues = 7;
#ifdef _DEBUG // Enables debug exceptions during optix launches. This may incur significant performance cost and should only be done during development.
pipeline_options.exceptionFlags =
OPTIX_EXCEPTION_FLAG_DEBUG | OPTIX_EXCEPTION_FLAG_TRACE_DEPTH | OPTIX_EXCEPTION_FLAG_STACK_OVERFLOW;
#else
pipeline_options.exceptionFlags = OPTIX_EXCEPTION_FLAG_NONE;
#endif
pipeline_options.pipelineLaunchParamsVariableName = "params";
pipeline_options.usesPrimitiveTypeFlags = OPTIX_PRIMITIVE_TYPE_FLAGS_CUSTOM;
const auto ptx = compile_cu2ptx("cuda/optix_curves_phantom.cu", "rayc_phantom", {CUDA_NVRTC_OPTIONS}, &compiler_log);
const size_t ptx_size = ptx.size();
if (!ptx_size)