forked from DJATOMs-archive/dfttest
-
Notifications
You must be signed in to change notification settings - Fork 0
/
dfttest.cpp
2287 lines (2190 loc) · 71.6 KB
/
dfttest.cpp
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
/*
** dfttest v1.9.4.3 for Avisynth+
**
** 2D/3D frequency domain denoiser.
**
** Copyright (C) 2007-2010 Kevin Stone, 2017 (C) DJATOM
**
** This program is free software; you can redistribute it and/or modify
** it under the terms of the GNU General Public License as published by
** the Free Software Foundation; either version 2 of the License, or
** (at your option) any later version.
**
** This program is distributed in the hope that it will be useful,
** but WITHOUT ANY WARRANTY; without even the implied warranty of
** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
** GNU General Public License for more details.
**
** You should have received a copy of the GNU General Public License
** along with this program; if not, write to the Free Software
** Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
/*
Modifications:
2018.10.14 - DJATOM
- Fixed one nasty bug, causing crash on non-AVX CPUs.
2017.09.04 - DJATOM
- Adaptive MT mode: MT_MULTI_INSTANCE for threads=1 and MT_SERIALIZED for > 1 internal
- Compilation: silence #693 for Intel Compiler
2017.08.14 - DJATOM
Changes from 1.9.4:
- x64 ready: ported almost all inline asm to intrinsics (dropped some SSE functions, we're in 2017 now).
- AddMean and RemoveMean got their AVX codepaths.
- PlanarFrame updated from JPSDR's NNEDI3 repo (x64 friendly, HBD ready).
- proc0_16 got SSE2 codepath (I can see noticeable speed-up against old version).
- opt parameter change: 2 - SSE/SSE2, 3 - SSE/SSE2/AVX codepath is used.
2009.12.25 - Firesledge
Added the bool "lsb" parameter. It outputs 16-bit pixel components by
separating most significant bytes (MSB) and least significant bytes (LSB).
When the flag is set, the output frame height is doubled. The top part of
the frame contains the MSB and the bottom part the LSB.
Default: false.
2011.05.15 - Firesledge
Added the bool "lsb_in" parameter to allow 16-bit input clips as stacked
MSB-LSB.
Default: false.
2011.11.28 - Firesledge
Added the quiet parameter to deactivate the filter spectrum output.
2012.03.11 - Firesledge
Fixed a stupid regression (from v1.8 mod16a) on the dither parameter.
2012.03.23 - Firesledge
The quiet parameter is not true by default.
2012.04.20 - Firesledge
Does no longer issue an error with null-length clips.
2013.07.31 - Firesledge
Compatible with the new Avisynth 2.6 colorspaces, excepted Y8.
*/
#ifdef __INTEL_COMPILER
#pragma warning(disable : 693)
#endif
#pragma warning(disable : 4305)
#include "dfttest.h"
#include "dfttest_avx.h"
#include <cassert>
PVideoFrame __stdcall dfttest::GetFrame(int n, IScriptEnvironment* env)
{
if (tbsize == 1)
return GetFrame_S(n, env);
return GetFrame_T(n, env);
}
unsigned __stdcall threadPool(void* ps)
{
const PS_INFO* pss = (PS_INFO*)ps;
while (true)
{
WaitForSingleObject(pss->nextJob, INFINITE);
if (pss->type < 0)
return 0;
if (!(pss->type & 2)) // tbsize=1
func_0(ps);
else
func_1(ps);
ResetEvent(pss->nextJob);
SetEvent(pss->jobFinished);
}
}
void proc0_C(const unsigned char* s0, const float* s1, float* d,
const int p0, const int p1, const int /*offset_lsb*/)
{
for (int u = 0; u < p1; ++u)
{
for (int v = 0; v < p1; ++v)
d[v] = s0[v] * s1[v];
s0 += p0;
s1 += p1;
d += p1;
}
}
void proc0_SSE2_4(const unsigned char* s0, const float* s1, float* d,
const int p0, const int p1, const int /*offset_lsb*/)
{
auto zero = _mm_setzero_si128();
for (int u = 0; u < p1; ++u)
{
for (int v = 0; v < p1; v += 4)
{
/* Intel's compiller works fine with aligned versions, MS randomly crashing */
auto s0_load = _mm_loadu_si128(reinterpret_cast<const __m128i*>(s0 + v));
auto s0_unp_lo1 = _mm_unpacklo_epi8(s0_load, zero);
auto s0_unp_lo2 = _mm_unpacklo_epi8(s0_unp_lo1, zero);
auto s0_loop = _mm_cvtepi32_ps(s0_unp_lo2);
auto s1_loop = _mm_loadu_ps(s1 + v);
auto d_reslt = _mm_mul_ps(s0_loop, s1_loop);
_mm_storeu_ps(d + v, d_reslt);
}
s0 += p0;
s1 += p1;
d += p1;
}
}
void proc0_SSE2_8(const unsigned char* s0, const float* s1, float* d,
const int p0, const int p1, const int /*offset_lsb*/)
{
auto zero = _mm_setzero_si128();
for (int u = 0; u < p1; ++u)
{
for (int v = 0; v < p1; v += 8)
{
auto s0_load = _mm_loadu_si128(reinterpret_cast<const __m128i*>(s0 + v));
auto s0_u_lo1 = _mm_unpacklo_epi8(s0_load, zero);
auto s0_u_lo2 = _mm_unpacklo_epi8(s0_u_lo1, zero);
auto s0_u_hi2 = _mm_unpackhi_epi8(s0_u_lo1, zero);
auto s0_loop_lo = _mm_cvtepi32_ps(s0_u_lo2);
auto s0_loop_hi = _mm_cvtepi32_ps(s0_u_hi2);
auto s1_loop_lo = _mm_loadu_ps(s1 + v);
auto s1_loop_hi = _mm_loadu_ps(s1 + v + 4);
auto d_result1 = _mm_mul_ps(s0_loop_lo, s1_loop_lo);
auto d_result2 = _mm_mul_ps(s0_loop_hi, s1_loop_hi);
_mm_storeu_ps(d + v, d_result1);
_mm_storeu_ps(d + v + 4, d_result2);
}
s0 += p0;
s1 += p1;
d += p1;
}
}
void proc0_16_C(const unsigned char* s0, const float* s1, float* d,
const int p0, const int p1, const int offset_lsb)
{
for (int u = 0; u < p1; ++u)
{
for (int v = 0; v < p1; ++v)
{
const int msb = s0[v];
const int lsb = s0[v + offset_lsb];
const int val = (msb << 8) + lsb;
d[v] = val * s1[v] * (1.0f / 256);
}
s0 += p0;
s1 += p1;
d += p1;
}
}
void proc0_16_SSE2(const unsigned char* s0, const float* s1, float* d,
const int p0, const int p1, const int offset_lsb)
{
auto zero = _mm_setzero_si128();
auto round = _mm_set_ps1(1.0f / 256);
for (int u = 0; u < p1; ++u)
{
for (int v = 0; v < p1; v += 4)
{
auto msb = _mm_loadu_si128(reinterpret_cast<const __m128i*>(s0 + v));
auto lsb = _mm_loadu_si128(reinterpret_cast<const __m128i*>(s0 + v + offset_lsb));
auto int_lo = _mm_unpacklo_epi8(lsb, msb);
auto int_hi = _mm_unpackhi_epi8(lsb, msb);
auto shift = _mm_unpacklo_epi8(_mm_slli_epi16(int_hi, 8), zero);
auto add = _mm_add_epi16(shift, int_lo);
auto s0_loop_i = _mm_unpacklo_epi16(add, zero);
auto s0_loop = _mm_cvtepi32_ps(s0_loop_i);
auto s1_loop = _mm_loadu_ps(s1 + v);
auto d_result = _mm_mul_ps(s0_loop, s1_loop);
d_result = _mm_mul_ps(d_result, round);
_mm_storeu_ps(d + v, d_result);
}
s0 += p0;
s1 += p1;
d += p1;
}
}
void proc1_C(const float* s0, const float* s1, float* d,
const int p0, const int p1)
{
for (int u = 0; u < p0; ++u)
{
for (int v = 0; v < p0; ++v)
d[v] += s0[v] * s1[v];
s0 += p0;
s1 += p0;
d += p1;
}
}
void proc1_SSE_4(const float* s0, const float* s1, float* d,
const int p0, const int p1)
{
auto zero = _mm_setzero_si128();
for (int u = 0; u < p0; ++u)
{
for (int v = 0; v < p0; v += 4)
{
auto s0_loop = _mm_loadu_ps(s0 + v);
auto s1_loop = _mm_loadu_ps(s1 + v);
auto d_loop = _mm_loadu_ps(d + v);
auto d_mul = _mm_mul_ps(s0_loop, s1_loop);
auto d_reslt = _mm_add_ps(d_loop, d_mul);
_mm_storeu_ps(d + v, d_reslt);
}
s0 += p0;
s1 += p0;
d += p1;
}
}
void proc1_SSE_8(const float* s0, const float* s1, float* d,
const int p0, const int p1)
{
for (int u = 0; u < p0; ++u)
{
for (int v = 0; v < p0; v += 8)
{
auto s0_loop1 = _mm_loadu_ps(s0 + v);
auto s0_loop2 = _mm_loadu_ps(s0 + v + 4);
auto s1_loop1 = _mm_loadu_ps(s1 + v);
auto s1_loop2 = _mm_loadu_ps(s1 + v + 4);
auto d_loop1 = _mm_loadu_ps(d + v);
auto d_loop2 = _mm_loadu_ps(d + v + 4);
auto d_mul1 = _mm_mul_ps(s0_loop1, s1_loop1);
auto d_mul2 = _mm_mul_ps(s0_loop2, s1_loop2);
auto d_reslt1 = _mm_add_ps(d_loop1, d_mul1);
auto d_reslt2 = _mm_add_ps(d_loop2, d_mul2);
_mm_storeu_ps(d + v, d_reslt1);
_mm_storeu_ps(d + v + 4, d_reslt2);
}
s0 += p0;
s1 += p0;
d += p1;
}
}
inline bool writeOk(const int* a, const int nthreads, const int x, const int y,
const int sbsize)
{
for (int i = 0; i < nthreads; ++i)
{
if (a[i * 2] != INT_MIN && abs(a[i * 2] - x) < sbsize &&
abs(a[i * 2 + 1] - y) < sbsize)
return false;
}
return true;
}
void func_0(void* ps)
{
PS_INFO* pss = (PS_INFO*)ps;
const int b = pss->b;
const int sheight = pss->sheight[b];
const int eheight = pss->eheight[b];
const int width = pss->pf->GetWidth(b);
const int src_pitch = pss->pf->GetPitch(b);
const unsigned char* srcp = pss->pf->GetPtr(b) + sheight * src_pitch;
float* ebp_saved = pss->ebuff[b] + width * sheight;
const int sbsize = pss->sbsize;
const int sbd1 = pss->sbsize >> 1;
const int sosize = pss->sosize;
LPCRITICAL_SECTION csect = pss->csect;
const int nthreads = pss->nthreads;
const int tidx = pss->tidx;
int* cty = pss->cty;
int* ctyT = cty + tidx * 2;
float* dftr = pss->dftr;
const float* hw = pss->hw;
fftwf_complex* dftgc = pss->dftgc;
fftwf_complex* dftc = pss->dftc;
fftwf_complex* dftc2 = pss->dftc2;
const bool zmean = pss->zmean;
const int ccnt = pss->ccnt;
const float f0beta = pss->f0beta;
const bool uf0b = fabsf(f0beta - 1.0f) < 0.00005f ? false : true;
const int inc = (pss->type & 1) ? sbsize - sosize : 1;
const int offset_lsb = pss->ofs_lsb[b];
for (int y = sheight; y < eheight; y += inc)
{
const bool tcheck = (nthreads > 1 && min(y - sheight, eheight - y) < sbsize) ? true : false;
for (int x = 0; x <= width - sbsize; x += inc)
{
pss->proc0(srcp + x, hw, dftr, src_pitch, sbsize, offset_lsb);
pss->fftwf_execute_dft_r2c(pss->ft, dftr, dftc);
if (zmean)
pss->removeMean((float*)dftc, (float*)dftgc, ccnt, (float*)dftc2);
pss->filterCoeffs((float*)dftc, pss->sigmas, ccnt,
uf0b ? &f0beta : pss->pmins, pss->pmaxs, pss->sigmas2);
if (zmean)
pss->addMean((float*)dftc, ccnt, (float*)dftc2);
pss->fftwf_execute_dft_c2r(pss->fti, dftc, dftr);
if (pss->type & 1) // spatial overlapping
{
if (tcheck) // potential thread conflict updating ebp
{
wloop:
EnterCriticalSection(csect);
// are any other threads updating pixels in this block?
if (writeOk(cty, nthreads, x, y, sbsize))
{
ctyT[0] = x; // if not, set that this thread is
ctyT[1] = y;
LeaveCriticalSection(csect);
}
else // if so, wait
{
LeaveCriticalSection(csect);
goto wloop;
}
}
pss->proc1(dftr, hw, ebp_saved + x, sbsize, width);
ctyT[0] = INT_MIN; // finished
}
else
ebp_saved[x + sbd1 * width + sbd1] =
dftr[sbd1 * sbsize + sbd1] * hw[sbd1 * sbsize + sbd1];
}
srcp += src_pitch * inc;
ebp_saved += width * inc;
}
}
void func_1(void* ps)
{
PS_INFO* pss = (PS_INFO*)ps;
nlCache* fc = pss->fc;
const int b = pss->b;
PlanarFrame* src = fc->frames[0]->ppf;
const int sheight = pss->sheight[b];
const int eheight = pss->eheight[b];
const int width = src->GetWidth(b);
const int src_pitch = src->GetPitch(b);
float** ebp = pss->ebuff;
const int sbsize = pss->sbsize;
const int sbd1 = pss->sbsize >> 1;
const int sosize = pss->sosize;
LPCRITICAL_SECTION csect = pss->csect;
const int nthreads = pss->nthreads;
const int tidx = pss->tidx;
int* cty = pss->cty;
int* ctyT = cty + tidx * 2;
const int stopz = pss->stopz;
const int barea = pss->barea;
const int pos = pss->pos;
float* dftr = pss->dftr;
const float* hw = pss->hw;
fftwf_complex* dftgc = pss->dftgc;
fftwf_complex* dftc = pss->dftc;
fftwf_complex* dftc2 = pss->dftc2;
const bool zmean = pss->zmean;
const int ccnt = pss->ccnt;
const float f0beta = pss->f0beta;
const bool uf0b = fabsf(f0beta - 1.0f) < 0.00005f ? false : true;
const unsigned char** pfplut = pss->pfplut;
for (int i = 0; i < fc->size; ++i)
pfplut[i] = fc->frames[fc->getCachePos(i)]->ppf->GetPtr(b) + src_pitch * sheight;
const int inc = (pss->type & 1) ? sbsize - sosize : 1;
const int offset_lsb = pss->ofs_lsb[b];
for (int y = sheight; y < eheight; y += inc)
{
const bool tcheck = (nthreads > 1 && min(y - sheight, eheight - y) < sbsize) ? true : false;
for (int x = 0; x <= width - sbsize; x += inc)
{
for (int z = 0; z <= stopz; ++z)
pss->proc0(pfplut[z] + x, hw + sbsize * sbsize * z,
dftr + sbsize * sbsize * z, src_pitch, sbsize, offset_lsb);
pss->fftwf_execute_dft_r2c(pss->ft, dftr, dftc);
if (zmean)
pss->removeMean((float*)dftc, (float*)dftgc, ccnt, (float*)dftc2);
pss->filterCoeffs((float*)dftc, pss->sigmas, ccnt,
uf0b ? &f0beta : pss->pmins, pss->pmaxs, pss->sigmas2);
if (zmean)
pss->addMean((float*)dftc, ccnt, (float*)dftc2);
pss->fftwf_execute_dft_c2r(pss->fti, dftc, dftr);
if (pss->type & 1) // spatial overlapping
{
if (tcheck) // potential thread conflict updating ebp
{
wloop:
EnterCriticalSection(csect);
// are any other threads updating pixels this block?
if (writeOk(cty, nthreads, x, y, sbsize))
{
ctyT[0] = x; // if not, set that this thread is
ctyT[1] = y;
LeaveCriticalSection(csect);
}
else // if so, wait
{
LeaveCriticalSection(csect);
goto wloop;
}
}
if (pss->type & 4) // temporal overlapping
{
for (int z = 0; z <= stopz; ++z)
{
pss->proc1(dftr + z * barea, hw + z * barea,
ebp[z * 3 + b] + y * width + x, sbsize, width);
}
}
else
{
pss->proc1(dftr + pos * barea, hw + pos * barea,
ebp[b] + y * width + x, sbsize, width);
}
ctyT[0] = INT_MIN; // finished
}
else
{
if (pss->type & 4) // temporal overlapping
{
for (int z = 0; z <= stopz; ++z)
ebp[z * 3 + b][(y + sbd1) * width + x + sbd1] +=
dftr[z * barea + sbd1 * sbsize + sbd1] * hw[z * barea + sbd1 * sbsize + sbd1];
}
else
{
ebp[b][(y + sbd1) * width + x + sbd1] =
dftr[pos * barea + sbd1 * sbsize + sbd1] * hw[pos * barea + sbd1 * sbsize + sbd1];
}
}
}
for (int q = 0; q < fc->size; ++q)
pfplut[q] += src_pitch * inc;
}
}
void intcast_C(const float* p, unsigned char* dst, const int src_height,
const int src_width, const int dst_pitch, const int width)
{
for (int y = 0; y < src_height; ++y)
{
for (int x = 0; x < src_width; ++x)
dst[x] = min(max((int)(p[x] + 0.5f), 0), 255);
p += width;
dst += dst_pitch;
}
}
void intcast_C_16_bits(const float* p, unsigned char* dst, unsigned char* dst_lsb, const int src_height,
const int src_width, const int dst_pitch, const int width)
{
const int saved_rounding_mode = ::_controlfp(RC_NEAR, _MCW_RC);
for (int y = 0; y < src_height; ++y)
{
for (int x = 0; x < src_width; ++x)
{
const float vf = p[x] * 256;
int v;
v = int(vf + 0.5f);
v = min(max(v, 0), 65535);
dst[x] = static_cast <unsigned char> (v >> 8);
dst_lsb[x] = static_cast <unsigned char> (v);
}
p += width;
dst += dst_pitch;
dst_lsb += dst_pitch;
}
::_controlfp(saved_rounding_mode, _MCW_RC);
}
void dither_C(const float* p, unsigned char* dst, const int src_height,
const int src_width, const int dst_pitch, const int width, const int mode)
{
float* dither = (float*)malloc(2 * width * sizeof(float));
float* dc = dither;
float* dn = dither + width;
const float scale = (mode - 1) + 0.5f;
const float off = scale * 0.5f;
MTRand mtr;
memset(dc, 0, width * sizeof(float));
for (int y = 0; y < src_height; ++y)
{
memset(dn, 0, width * sizeof(float));
for (int x = 0; x < src_width; ++x)
{
const int v = mode == 1 ?
(int)(p[x] + dc[x] + 0.5f) :
(int)(p[x] + mtr.randf() * scale - off + dc[x] + 0.5f);
dst[x] = min(max(v, 0), 255);
const float qerror = p[x] - dst[x];
if (x != 0)
dn[x - 1] += qerror * 0.1875f;
dn[x] += qerror * 0.3125f;
if (x != src_width - 1)
{
dc[x + 1] += qerror * 0.4375f;
dn[x + 1] += qerror * 0.0625f;
}
}
p += width;
dst += dst_pitch;
float* tn = dn;
dn = dc;
dc = tn;
}
free(dither);
}
void intcast_SSE2_8(const float* p, unsigned char* dst, const int src_height,
const int src_width, const int dst_pitch, const int width)
{
auto sse2_05 = _mm_set_ps1(0.5f);
auto sse2_255 = _mm_set_ps1(255);
auto sse2_0 = _mm_set_ps1(0);
for (int y = 0; y < src_height; ++y)
{
for (int x = 0; x < src_width; x += 8)
{
auto p_loop = _mm_loadu_ps(p + x);
auto p_loop2 = _mm_loadu_ps(p + x + 4);
auto add_loop = _mm_add_ps(sse2_05, p_loop);
auto add_loop2 = _mm_add_ps(sse2_05, p_loop2);
auto max_loop = _mm_max_ps(add_loop, sse2_0);
auto max_loop2 = _mm_max_ps(add_loop2, sse2_0);
auto min_loop = _mm_min_ps(max_loop, sse2_255);
auto min_loop2 = _mm_min_ps(max_loop2, sse2_255);
auto int1_loop = _mm_cvttps_epi32(min_loop);
auto int1_loop2 = _mm_cvttps_epi32(min_loop2);
auto packs_loop = _mm_packs_epi32(int1_loop, int1_loop2);
auto result = _mm_packus_epi16(packs_loop, packs_loop);
_mm_storeu_si128(reinterpret_cast<__m128i*>(dst + x), result);
}
p += width;
dst += dst_pitch;
}
}
PVideoFrame dfttest::GetFrame_S(int n, IScriptEnvironment* env)
{
PVideoFrame src = child->GetFrame(mapn(n), env);
nlf->pf->copyFrom(src, vi_src);
copyPad(nlf->pf, nlf->ppf, env);
for (int b = 0; b < 3; ++b)
{
if ((b == 0 && !Y) || (b == 1 && !U) || (b == 2 && !V))
{
bypass_plane(*(nlf->pf), b);
continue;
}
const int height = nlf->ppf->GetHeight(b) / lsb_in_hmul;
const int width = nlf->ppf->GetWidth(b);
memset(ebuff[b], 0, height * width * sizeof(float));
for (int i = 0; i < threads; ++i)
{
pssInfo[i]->b = b;
ResetEvent(pssInfo[i]->jobFinished);
SetEvent(pssInfo[i]->nextJob);
}
for (int i = 0; i < threads; ++i)
WaitForSingleObject(pssInfo[i]->jobFinished, INFINITE);
conv_result_plane_to_int(width, height, b, b, env);
}
return build_output_frame(env);
}
PVideoFrame dfttest::GetFrame_T(int n, IScriptEnvironment* env)
{
int pos = 0;
PlanarFrame* src = fc->frames[0]->ppf;
if (tmode == 0)
{
for (int b = 0; b < 3; ++b)
{
if ((b == 0 && !Y) || (b == 1 && !U) || (b == 2 && !V))
continue;
memset(ebuff[b], 0, (src->GetHeight(b) / lsb_in_hmul) * src->GetWidth(b) * sizeof(float));
}
pos = tbsize >> 1;
fc->resetCacheStart(n - pos, n + pos);
for (int i = n - pos; i <= n + pos; ++i)
{
nlFrame* nl = fc->frames[fc->getCachePos(i - n + pos)];
if (nl->fnum != i)
{
PVideoFrame src = child->GetFrame(mapn(i), env);
nl->pf->copyFrom(src, vi_src);
copyPad(nl->pf, nl->ppf, env);
nl->setFNum(i);
}
}
processTemporalBlock(pos);
}
else
{
int fbframe = -max((tbsize - tosize), tosize);
for (; fbframe <= n - tbsize; fbframe += (tbsize - tosize));
int lbframe = fbframe;
for (; lbframe <= n - (tbsize - tosize); lbframe += (tbsize - tosize));
if (ebframe >= fbframe && ebframe <= lbframe) // cached
{
pos = n - ebframe;
fbframe = ebframe + (tbsize - tosize);
}
else
{
fbframe = -max((tbsize - tosize), tosize);
for (; fbframe <= lbframe - tbsize; fbframe += (tbsize - tosize));
for (int q = tbsize - tosize; q < tbsize; ++q)
{
for (int b = 0; b < 3; ++b)
{
if ((b == 0 && !Y) || (b == 1 && !U) || (b == 2 && !V))
continue;
memset(ebuff[q * 3 + b], 0, (src->GetHeight(b) / lsb_in_hmul) * src->GetWidth(b) * sizeof(float));
}
}
}
for (int z = fbframe; z <= lbframe; z += (tbsize - tosize))
{
for (int q = 0; q < tbsize - tosize; ++q)
{
float* tempb[3] = { ebuff[0], ebuff[1], ebuff[2] };
for (int k = 0; k < tbsize - 1; ++k)
{
ebuff[k * 3 + 0] = ebuff[(k + 1) * 3 + 0];
ebuff[k * 3 + 1] = ebuff[(k + 1) * 3 + 1];
ebuff[k * 3 + 2] = ebuff[(k + 1) * 3 + 2];
}
ebuff[(tbsize - 1) * 3 + 0] = tempb[0];
ebuff[(tbsize - 1) * 3 + 1] = tempb[1];
ebuff[(tbsize - 1) * 3 + 2] = tempb[2];
}
for (int q = tosize; q < tbsize; ++q)
{
for (int b = 0; b < 3; ++b)
{
if ((b == 0 && !Y) || (b == 1 && !U) || (b == 2 && !V))
continue;
memset(ebuff[q * 3 + b], 0, (src->GetHeight(b) / lsb_in_hmul) * src->GetWidth(b) * sizeof(float));
}
}
pos = n - z;
fc->resetCacheStart(z, z + tbsize - 1);
for (int i = z; i <= z + tbsize - 1; ++i)
{
nlFrame* nl = fc->frames[fc->getCachePos(i - z)];
if (nl->fnum != i)
{
PVideoFrame src = child->GetFrame(mapn(i), env);
nl->pf->copyFrom(src, vi_src);
copyPad(nl->pf, nl->ppf, env);
nl->setFNum(i);
}
}
processTemporalBlock(pos);
}
ebframe = lbframe;
}
for (int b = 0; b < 3; ++b)
{
if ((b == 0 && !Y) || (b == 1 && !U) || (b == 2 && !V))
{
bypass_plane(*(fc->frames[fc->getCachePos(pos)]->pf), b);
continue;
}
const int height = src->GetHeight(b) / lsb_in_hmul;
const int width = src->GetWidth(b);
conv_result_plane_to_int(width, height, b, tmode == 0 ? b : (pos * 3 + b), env);
}
return build_output_frame(env);
}
PVideoFrame dfttest::build_output_frame(IScriptEnvironment* env)
{
PVideoFrame dst = env->NewVideoFrame(vi);
if (lsb_out_flag)
{
merge_msb_lsb();
dstPF_all->copyTo(dst, vi);
}
else
{
dstPF->copyTo(dst, vi);
}
return dst;
}
void dfttest::bypass_plane(PlanarFrame& frame, int plane)
{
if (lsb_in_flag)
{
if (lsb_out_flag)
{
frame.copyPlaneTo16Part(*dstPF, plane, 0);
frame.copyPlaneTo16Part(*dstPF_lsb, plane, 1);
}
else
{
frame.copyPlaneTo16Round(*dstPF, plane);
}
}
else
{
frame.copyPlaneTo(*dstPF, plane);
if (lsb_out_flag)
{
unsigned char* plane_ptr = dstPF_lsb->GetPtr(plane);
const int dp = dstPF_lsb->GetPitch(plane);
const int dh = dstPF_lsb->GetHeight(plane);
memset(plane_ptr, 0, dp * dh);
}
}
}
void dfttest::conv_result_plane_to_int(int width, int height, int b, int ebuff_index, IScriptEnvironment* env)
{
unsigned char* dstp = dstPF->GetPtr(b);
const int dst_pitch = dstPF->GetPitch(b);
const int src_width = dstPF->GetWidth(b);
const int src_height = dstPF->GetHeight(b);
float* ebp = ebuff[ebuff_index] + width * ((height - src_height) >> 1) + ((width - src_width) >> 1);
long cpuflags = env->GetCPUFlags();
if (lsb_out_flag)
{
assert(dstPF_lsb != 0);
unsigned char* dst_lsb_p = dstPF_lsb->GetPtr(b);
intcast_C_16_bits(
ebp,
dstp, dst_lsb_p,
src_height, src_width,
dst_pitch, width
);
}
else
{
if (dither)
dither_C(ebp, dstp, src_height, src_width, dst_pitch, width, dither);
else if (!(src_width & 7) && (((cpuflags & CPUF_SSE2) && opt == 0) || opt == 3 || opt == 2))
intcast_SSE2_8(ebp, dstp, src_height, src_width, dst_pitch, width);
else
intcast_C(ebp, dstp, src_height, src_width, dst_pitch, width);
}
}
void dfttest::merge_msb_lsb()
{
for (int b = 0; b < 3; ++b)
{
const int width = dstPF->GetWidth(b);
const int height = dstPF->GetHeight(b);
const int msb_pitch = dstPF->GetPitch(b);
const unsigned char* msb_ptr = dstPF->GetPtr(b);
const int lsb_pitch = dstPF_lsb->GetPitch(b);
const unsigned char* lsb_ptr = dstPF_lsb->GetPtr(b);
const int all_pitch = dstPF_all->GetPitch(b);
unsigned char* all_ptr = dstPF_all->GetPtr(b);
for (int y = 0; y < height; ++y)
{
memcpy(
all_ptr + y * all_pitch,
msb_ptr + y * msb_pitch,
width
);
memcpy(
all_ptr + (height + y) * all_pitch,
lsb_ptr + y * lsb_pitch,
width
);
}
}
}
void dfttest::processTemporalBlock(int pos)
{
for (int i = 0; i < threads; ++i)
pssInfo[i]->pos = pos;
for (int b = 0; b < 3; ++b)
{
if ((b == 0 && !Y) || (b == 1 && !U) || (b == 2 && !V))
continue;
for (int i = 0; i < threads; ++i)
{
pssInfo[i]->b = b;
ResetEvent(pssInfo[i]->jobFinished);
SetEvent(pssInfo[i]->nextJob);
}
for (int i = 0; i < threads; ++i)
WaitForSingleObject(pssInfo[i]->jobFinished, INFINITE);
}
}
void removeMean_C(float* dftc, const float* dftgc, const int ccnt, float* dftc2)
{
const float gf = dftc[0] / dftgc[0];
for (int h = 0; h < ccnt; h += 2)
{
dftc2[h + 0] = gf * dftgc[h + 0];
dftc2[h + 1] = gf * dftgc[h + 1];
dftc[h + 0] -= dftc2[h + 0];
dftc[h + 1] -= dftc2[h + 1];
}
}
void removeMean_SSE(float* dftc, const float* dftgc, const int ccnt, float* dftc2)
{
const float gf = dftc[0] / dftgc[0];
auto gf_asm = _mm_load_ps1(reinterpret_cast<const float*>(&gf));
for (int h = 0; h < ccnt; h += 4)
{
auto dftc_loop = _mm_loadu_ps(dftc + h);
auto dftgc_loop = _mm_loadu_ps(dftgc + h);
auto dftc2_loop = _mm_loadu_ps(dftc2 + h);
auto dftc2_result = _mm_mul_ps(gf_asm, dftgc_loop);
auto dftc_result = _mm_sub_ps(dftc_loop, dftc2_result);
_mm_storeu_ps(dftc2 + h, dftc2_result);
_mm_storeu_ps(dftc + h, dftc_result);
}
}
void addMean_C(float* dftc, const int ccnt, const float* dftc2)
{
for (int h = 0; h < ccnt; h += 2)
{
dftc[h + 0] += dftc2[h + 0];
dftc[h + 1] += dftc2[h + 1];
}
}
void addMean_SSE(float* dftc, const int ccnt, const float* dftc2)
{
for (int h = 0; h < ccnt; h += 4)
{
auto dftc_loop = _mm_loadu_ps(dftc + h);
auto dftc2_loop = _mm_loadu_ps(dftc2 + h);
auto dftc_result = _mm_add_ps(dftc2_loop, dftc_loop);
_mm_storeu_ps(dftc + h, dftc_result);
}
}
void filter_0_C(float* dftc, const float* sigmas, const int ccnt,
const float* pmin, const float* pmax, const float* sigmas2)
{
for (int h = 0; h < ccnt; h += 2)
{
const float psd = dftc[h + 0] * dftc[h + 0] + dftc[h + 1] * dftc[h + 1];
const float coeff = max((psd - sigmas[h]) / (psd + 1e-15f), 0.0f);
dftc[h + 0] *= coeff;
dftc[h + 1] *= coeff;
}
}
void filter_0_SSE(float* dftc, const float* sigmas, const int ccnt,
const float* pmin, const float* pmax, const float* sigmas2)
{
auto zero = _mm_setzero_ps();
auto sse_1em15 = _mm_set_ps1(1e-15f);
for (int h = 0; h < ccnt; h += 4)
{
auto dftc_loop = _mm_loadu_ps(dftc + h);
auto sigmas_loop = _mm_loadu_ps(sigmas + h);
auto mul1_loop = _mm_mul_ps(dftc_loop, dftc_loop);
auto shuff_loop = _mm_shuffle_ps(mul1_loop, mul1_loop, 177);
auto add1_loop = _mm_add_ps(mul1_loop, shuff_loop);
auto sub1_loop = _mm_sub_ps(add1_loop, sigmas_loop);
auto add2_loop = _mm_add_ps(add1_loop, sse_1em15);
auto rcp1_loop = _mm_rcp_ps(add2_loop);
auto mul2_loop = _mm_mul_ps(sub1_loop, rcp1_loop);
auto max1_loop = _mm_max_ps(zero, mul2_loop);
auto mul3_loop = _mm_mul_ps(dftc_loop, max1_loop);
_mm_storeu_ps(dftc + h, mul3_loop);
}
}
void filter_1_C(float* dftc, const float* sigmas, const int ccnt,
const float* pmin, const float* pmax, const float* sigmas2)
{
for (int h = 0; h < ccnt; h += 2)
{
const float psd = dftc[h + 0] * dftc[h + 0] + dftc[h + 1] * dftc[h + 1];
if (psd < sigmas[h])
dftc[h + 0] = dftc[h + 1] = 0.0f;
}
}
void filter_1_SSE(float* dftc, const float* sigmas, const int ccnt,
const float* pmin, const float* pmax, const float* sigmas2)
{
for (int h = 0; h < ccnt; h += 4)
{
auto dftc_loop = _mm_loadu_ps(dftc + h);
auto sigmas_loop = _mm_loadu_ps(sigmas + h);
auto mul1_loop = _mm_mul_ps(dftc_loop, dftc_loop);
auto shuff_loop = _mm_shuffle_ps(mul1_loop, mul1_loop, 177);
auto add1_loop = _mm_add_ps(mul1_loop, shuff_loop);
auto cmple1_loop = _mm_cmple_ps(sigmas_loop, add1_loop);
auto and1_loop = _mm_and_ps(cmple1_loop, dftc_loop);
_mm_storeu_ps(dftc + h, and1_loop);
}
}
void filter_2_C(float* dftc, const float* sigmas, const int ccnt,
const float* pmin, const float* pmax, const float* sigmas2)
{
for (int h = 0; h < ccnt; h += 2)
{
dftc[h + 0] *= sigmas[h];
dftc[h + 1] *= sigmas[h];
}
}
void filter_2_SSE(float* dftc, const float* sigmas, const int ccnt,
const float* pmin, const float* pmax, const float* sigmas2)
{
for (int h = 0; h < ccnt; h += 4)
{
auto dftc_loop = _mm_loadu_ps(dftc + h);
auto sigmas_loop = _mm_loadu_ps(sigmas + h);
auto mul_loop = _mm_mul_ps(dftc_loop, sigmas_loop);
_mm_storeu_ps(dftc + h, mul_loop);
}
}
void filter_3_C(float* dftc, const float* sigmas, const int ccnt,
const float* pmin, const float* pmax, const float* sigmas2)
{
for (int h = 0; h < ccnt; h += 2)
{
const float psd = dftc[h + 0] * dftc[h + 0] + dftc[h + 1] * dftc[h + 1];
if (psd >= pmin[h] && psd <= pmax[h])
{
dftc[h + 0] *= sigmas[h];
dftc[h + 1] *= sigmas[h];
}
else
{
dftc[h + 0] *= sigmas2[h];
dftc[h + 1] *= sigmas2[h];
}
}
}
void filter_3_SSE(float* dftc, const float* sigmas, const int ccnt,
const float* pmin, const float* pmax, const float* sigmas2)
{
auto sse_ones = _mm_set_ps1(0xFFFFFFFFFFFFFFFF);
for (int h = 0; h < ccnt; h += 4)
{
auto dftc_loop = _mm_loadu_ps(dftc + h);
auto sigmas_loop = _mm_loadu_ps(sigmas + h);
auto sigmas2_loop = _mm_loadu_ps(sigmas2 + h);
auto pmin_loop = _mm_loadu_ps(pmin + h);
auto pmax_loop = _mm_loadu_ps(pmax + h);
auto mul1_loop = _mm_mul_ps(dftc_loop, dftc_loop);
auto shuff_loop = _mm_shuffle_ps(mul1_loop, mul1_loop, 177);
auto add1_loop = _mm_add_ps(mul1_loop, shuff_loop);
auto cmpnle1_loop = _mm_cmpnle_ps(pmax_loop, add1_loop);
auto cmple1_loop = _mm_cmple_ps(pmin_loop, add1_loop);
auto and1_loop = _mm_and_ps(cmple1_loop, cmpnle1_loop);
auto and2_loop = _mm_and_ps(sigmas_loop, and1_loop);
auto xor1_loop = _mm_xor_ps(sse_ones, and1_loop);
auto and3_loop = _mm_and_ps(xor1_loop, sigmas2_loop);
auto or1_loop = _mm_or_ps(and3_loop, and2_loop);
auto mul2_loop = _mm_mul_ps(or1_loop, dftc_loop);
_mm_storeu_ps(dftc + h, mul2_loop);
}
}
void filter_4_C(float* dftc, const float* sigmas, const int ccnt,