-
Notifications
You must be signed in to change notification settings - Fork 5
/
Tensor_hip.cpp
1693 lines (1462 loc) · 73.9 KB
/
Tensor_hip.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
/*
MIT License
Copyright (c) 2019 - 2024 Advanced Micro Devices, Inc.
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
*/
#include <stdio.h>
#include <dirent.h>
#include <string.h>
#include <opencv2/core/core.hpp>
#include <opencv2/highgui/highgui.hpp>
#include <opencv2/opencv.hpp>
#include <iostream>
#include "rpp.h"
#include "../rpp_test_suite_common.h"
#include <sys/types.h>
#include <sys/stat.h>
#include <unistd.h>
#include <time.h>
#include <omp.h>
#include <hip/hip_fp16.h>
#include <fstream>
typedef half Rpp16f;
using namespace cv;
using namespace std;
int main(int argc, char **argv)
{
// Handle inputs
const int MIN_ARG_COUNT = 19;
char *src = argv[1];
char *srcSecond = argv[2];
string dst = argv[3];
int inputBitDepth = atoi(argv[4]);
unsigned int outputFormatToggle = atoi(argv[5]);
int testCase = atoi(argv[6]);
int numRuns = atoi(argv[8]);
int testType = atoi(argv[9]); // 0 for unit and 1 for performance test
int layoutType = atoi(argv[10]); // 0 for pkd3 / 1 for pln3 / 2 for pln1
int qaFlag = atoi(argv[12]);
int decoderType = atoi(argv[13]);
int batchSize = atoi(argv[14]);
bool additionalParamCase = (testCase == 8 || testCase == 21 || testCase == 23|| testCase == 24 || testCase == 40 || testCase == 41 || testCase == 49 || testCase == 54 || testCase == 79);
bool kernelSizeCase = (testCase == 40 || testCase == 41 || testCase == 49 || testCase == 54);
bool dualInputCase = (testCase == 2 || testCase == 30 || testCase == 33 || testCase == 61 || testCase == 63 || testCase == 65 || testCase == 68);
bool randomOutputCase = (testCase == 6 || testCase == 8 || testCase == 84 || testCase == 49 || testCase == 54);
bool nonQACase = (testCase == 24);
bool interpolationTypeCase = (testCase == 21 || testCase == 23 || testCase == 24 || testCase == 79);
bool reductionTypeCase = (testCase == 87 || testCase == 88 || testCase == 89 || testCase == 90 || testCase == 91);
bool noiseTypeCase = (testCase == 8);
bool pln1OutTypeCase = (testCase == 86);
unsigned int verbosity = atoi(argv[11]);
unsigned int additionalParam = additionalParamCase ? atoi(argv[7]) : 1;
int roiList[4] = {atoi(argv[15]), atoi(argv[16]), atoi(argv[17]), atoi(argv[18])};
string scriptPath = argv[19];
if (verbosity == 1)
{
cout << "\nInputs for this test case are:";
cout << "\nsrc1 = " << argv[1];
cout << "\nsrc2 = " << argv[2];
if (testType == 0)
cout << "\ndst = " << argv[3];
cout << "\nu8 / f16 / f32 / u8->f16 / u8->f32 / i8 / u8->i8 (0/1/2/3/4/5/6) = " << argv[4];
cout << "\noutputFormatToggle (pkd->pkd = 0 / pkd->pln = 1) = " << argv[5];
cout << "\ncase number (0:91) = " << argv[6];
cout << "\nnumber of times to run = " << argv[8];
cout << "\ntest type - (0 = unit tests / 1 = performance tests) = " << argv[9];
cout << "\nlayout type - (0 = PKD3/ 1 = PLN3/ 2 = PLN1) = " << argv[10];
cout << "\nqa mode - 0/1 = " << argv[12];
cout << "\ndecoder type - (0 = TurboJPEG / 1 = OpenCV) = " << argv[13];
cout << "\nbatch size = " << argv[14];
}
if (argc < MIN_ARG_COUNT)
{
cout << "\nImproper Usage! Needs all arguments!\n";
cout << "\nUsage: <src1 folder> <src2 folder (place same as src1 folder for single image functionalities)> <dst folder> <u8 = 0 / f16 = 1 / f32 = 2 / u8->f16 = 3 / u8->f32 = 4 / i8 = 5 / u8->i8 = 6> <outputFormatToggle (pkd->pkd = 0 / pkd->pln = 1)> <case number = 0:87> <number of runs > 0> <layout type (0 = PKD3/ 1 = PLN3/ 2 = PLN1)> <qa mode (0/1)> <decoder type (0/1)> <batch size > 1> <roiList> <verbosity = 0/1>>\n";
return -1;
}
if (layoutType == 2)
{
if(testCase == 36 || testCase == 31 || testCase == 35 || testCase == 45 || testCase == 86)
{
cout << "\ncase " << testCase << " does not exist for PLN1 layout\n";
return -1;
}
else if (outputFormatToggle != 0)
{
cout << "\nPLN1 cases don't have outputFormatToggle! Please input outputFormatToggle = 0\n";
return -1;
}
}
if(pln1OutTypeCase && outputFormatToggle != 0)
{
cout << "\ntest case " << testCase << " don't have outputFormatToggle! Please input outputFormatToggle = 0\n";
return -1;
}
else if (reductionTypeCase && outputFormatToggle != 0)
{
cout << "\nReduction Kernels don't have outputFormatToggle! Please input outputFormatToggle = 0\n";
return -1;
}
else if(batchSize > MAX_BATCH_SIZE)
{
std::cerr << "\n Batchsize should be less than or equal to "<< MAX_BATCH_SIZE << " Aborting!";
exit(0);
}
else if(testCase == 82 && batchSize < 2)
{
std::cerr<<"\n RICAP only works with BatchSize > 1";
exit(0);
}
// Get function name
string funcName = augmentationMap[testCase];
if (funcName.empty())
{
if (testType == 0)
cout << "\ncase " << testCase << " is not supported\n";
return -1;
}
// Determine the number of input channels based on the specified layout type
int inputChannels = set_input_channels(layoutType);
// Determine the type of function to be used based on the specified layout type
string funcType = set_function_type(layoutType, pln1OutTypeCase, outputFormatToggle, "HIP");
// Initialize tensor descriptors
RpptDesc srcDesc, dstDesc;
RpptDescPtr srcDescPtr = &srcDesc;
RpptDescPtr dstDescPtr = &dstDesc;
// Set src/dst layout types in tensor descriptors
set_descriptor_layout( srcDescPtr, dstDescPtr, layoutType, pln1OutTypeCase, outputFormatToggle);
// Set src/dst data types in tensor descriptors
set_descriptor_data_type(inputBitDepth, funcName, srcDescPtr, dstDescPtr);
// Other initializations
int missingFuncFlag = 0;
int i = 0, j = 0;
int maxHeight = 0, maxWidth = 0;
int maxDstHeight = 0, maxDstWidth = 0;
Rpp64u count = 0;
Rpp64u ioBufferSize = 0;
Rpp64u oBufferSize = 0;
static int noOfImages = 0;
Mat image, imageSecond;
// String ops on input path
string inputPath = src;
inputPath += "/";
string inputPathSecond = srcSecond;
inputPathSecond += "/";
string func = funcName;
func += funcType;
RpptInterpolationType interpolationType = RpptInterpolationType::BILINEAR;
std::string interpolationTypeName = "";
std::string noiseTypeName = "";
if (kernelSizeCase)
{
char additionalParam_char[2];
std::sprintf(additionalParam_char, "%u", additionalParam);
func += "_kSize";
func += additionalParam_char;
}
else if (interpolationTypeCase)
{
interpolationTypeName = get_interpolation_type(additionalParam, interpolationType);
func += "_interpolationType";
func += interpolationTypeName.c_str();
}
else if (noiseTypeCase)
{
noiseTypeName = get_noise_type(additionalParam);
func += "_noiseType";
func += noiseTypeName.c_str();
}
if(!qaFlag)
{
dst += "/";
dst += func;
}
// Get number of images and image Names
vector<string> imageNames, imageNamesSecond, imageNamesPath, imageNamesPathSecond;
search_files_recursive(src, imageNames, imageNamesPath, ".jpg");
if(dualInputCase)
{
search_files_recursive(srcSecond, imageNamesSecond, imageNamesPathSecond, ".jpg");
if(imageNames.size() != imageNamesSecond.size())
{
std::cerr << " \n The number of images in the input folders must be the same.";
exit(0);
}
}
noOfImages = imageNames.size();
if(noOfImages < batchSize || ((noOfImages % batchSize) != 0))
{
replicate_last_file_to_fill_batch(imageNamesPath[noOfImages - 1], imageNamesPath, imageNames, imageNames[noOfImages - 1], noOfImages, batchSize);
if(dualInputCase)
replicate_last_file_to_fill_batch(imageNamesPathSecond[noOfImages - 1], imageNamesPathSecond, imageNamesSecond, imageNamesSecond[noOfImages - 1], noOfImages, batchSize);
noOfImages = imageNames.size();
}
if(!noOfImages)
{
std::cerr << "Not able to find any images in the folder specified. Please check the input path";
exit(0);
}
if(qaFlag)
{
sort(imageNames.begin(), imageNames.end());
if(dualInputCase)
sort(imageNamesSecond.begin(), imageNamesSecond.end());
}
// Initialize ROI tensors for src/dst
RpptROI *roiTensorPtrSrc, *roiTensorPtrDst;
CHECK_RETURN_STATUS(hipHostMalloc(&roiTensorPtrSrc, batchSize * sizeof(RpptROI)));
CHECK_RETURN_STATUS(hipHostMalloc(&roiTensorPtrDst, batchSize * sizeof(RpptROI)));
// Initialize the ImagePatch for dst
RpptImagePatch *dstImgSizes;
CHECK_RETURN_STATUS(hipHostMalloc(&dstImgSizes, batchSize * sizeof(RpptImagePatch)));
// Set ROI tensors types for src/dst
RpptRoiType roiTypeSrc, roiTypeDst;
roiTypeSrc = RpptRoiType::XYWH;
roiTypeDst = RpptRoiType::XYWH;
Rpp32u outputChannels = inputChannels;
if(pln1OutTypeCase)
outputChannels = 1;
Rpp32u srcOffsetInBytes = (kernelSizeCase) ? (12 * (additionalParam / 2)) : 0;
Rpp32u dstOffsetInBytes = 0;
int imagesMixed = 0; // Flag used to check if all images in dataset is of same dimensions
set_max_dimensions(imageNamesPath, maxHeight, maxWidth, imagesMixed);
if(testCase == 82 && imagesMixed)
{
std::cerr<<"\n RICAP only works with same dimension images";
exit(0);
}
// Set numDims, offset, n/c/h/w values, strides for src/dst
set_descriptor_dims_and_strides(srcDescPtr, batchSize, maxHeight, maxWidth, inputChannels, srcOffsetInBytes);
set_descriptor_dims_and_strides(dstDescPtr, batchSize, maxHeight, maxWidth, outputChannels, dstOffsetInBytes);
// Factors to convert U8 data to F32, F16 data to 0-1 range and reconvert them back to 0 -255 range
Rpp32f conversionFactor = 1.0f / 255.0;
if(testCase == 38)
conversionFactor = 1.0;
Rpp32f invConversionFactor = 1.0f / conversionFactor;
// Set buffer sizes in pixels for src/dst
ioBufferSize = (Rpp64u)srcDescPtr->h * (Rpp64u)srcDescPtr->w * (Rpp64u)srcDescPtr->c * (Rpp64u)batchSize;
oBufferSize = (Rpp64u)dstDescPtr->h * (Rpp64u)dstDescPtr->w * (Rpp64u)dstDescPtr->c * (Rpp64u)batchSize;
// Set buffer sizes in bytes for src/dst (including offsets)
Rpp64u ioBufferSizeInBytes_u8 = ioBufferSize + srcDescPtr->offsetInBytes;
Rpp64u oBufferSizeInBytes_u8 = oBufferSize + dstDescPtr->offsetInBytes;
Rpp64u inputBufferSize = ioBufferSize * get_size_of_data_type(srcDescPtr->dataType) + srcDescPtr->offsetInBytes;
Rpp64u outputBufferSize = oBufferSize * get_size_of_data_type(dstDescPtr->dataType) + dstDescPtr->offsetInBytes;
// Initialize 8u host buffers for src/dst
Rpp8u *inputu8 = static_cast<Rpp8u *>(calloc(ioBufferSizeInBytes_u8, 1));
Rpp8u *inputu8Second = static_cast<Rpp8u *>(calloc(ioBufferSizeInBytes_u8, 1));
Rpp8u *outputu8 = static_cast<Rpp8u *>(calloc(oBufferSizeInBytes_u8, 1));
if (testCase == 40) memset(inputu8, 0xFF, ioBufferSizeInBytes_u8);
Rpp8u *offsettedInput, *offsettedInputSecond;
offsettedInput = inputu8 + srcDescPtr->offsetInBytes;
offsettedInputSecond = inputu8Second + srcDescPtr->offsetInBytes;
void *input, *input_second, *output;
void *d_input, *d_input_second, *d_output;
input = static_cast<Rpp8u *>(calloc(inputBufferSize, 1));
input_second = static_cast<Rpp8u *>(calloc(inputBufferSize, 1));
output = static_cast<Rpp8u *>(calloc(outputBufferSize, 1));
Rpp32f *rowRemapTable, *colRemapTable;
if(testCase == 79)
{
rowRemapTable = static_cast<Rpp32f *>(calloc(ioBufferSize, sizeof(Rpp32f)));
colRemapTable = static_cast<Rpp32f *>(calloc(ioBufferSize, sizeof(Rpp32f)));
}
// Run case-wise RPP API and measure time
rppHandle_t handle;
hipStream_t stream;
CHECK_RETURN_STATUS(hipStreamCreate(&stream));
rppCreateWithStreamAndBatchSize(&handle, stream, batchSize);
int noOfIterations = (int)imageNames.size() / batchSize;
double maxWallTime = 0, minWallTime = 500, avgWallTime = 0;
double wallTime;
string testCaseName;
// Initialize buffers for any reductionType functions (testCase 87 - tensor_sum alone cannot return final sum as 8u/8s due to overflow. 8u inputs return 64u sums, 8s inputs return 64s sums)
void *reductionFuncResultArr;
Rpp32f *mean;
Rpp32u reductionFuncResultArrLength = srcDescPtr->n * 4;
if (reductionTypeCase)
{
int bitDepthByteSize = 0;
if ((dstDescPtr->dataType == RpptDataType::F16) || (dstDescPtr->dataType == RpptDataType::F32) || testCase == 90 || testCase == 91)
bitDepthByteSize = sizeof(Rpp32f); // using 32f outputs for 16f and 32f, for testCase 90, 91
else if ((dstDescPtr->dataType == RpptDataType::U8) || (dstDescPtr->dataType == RpptDataType::I8))
bitDepthByteSize = (testCase == 87) ? sizeof(Rpp64u) : sizeof(Rpp8u);
CHECK_RETURN_STATUS(hipHostMalloc(&reductionFuncResultArr, reductionFuncResultArrLength * bitDepthByteSize));
if(testCase == 91)
CHECK_RETURN_STATUS(hipHostMalloc(&mean, reductionFuncResultArrLength * bitDepthByteSize));
}
// create generic descriptor and params in case of slice
RpptGenericDesc descriptor3D;
RpptGenericDescPtr descriptorPtr3D = &descriptor3D;
Rpp32s *anchorTensor = NULL, *shapeTensor = NULL;
Rpp32u *roiTensor = NULL;
if(testCase == 92)
set_generic_descriptor_slice(srcDescPtr, descriptorPtr3D, batchSize);
// Allocate hip memory for src/dst
CHECK_RETURN_STATUS(hipMalloc(&d_input, inputBufferSize));
CHECK_RETURN_STATUS(hipMalloc(&d_output, outputBufferSize));
if(dualInputCase)
CHECK_RETURN_STATUS(hipMalloc(&d_input_second, inputBufferSize));
RpptROI *roiPtrInputCropRegion;
if(testCase == 82)
CHECK_RETURN_STATUS(hipHostMalloc(&roiPtrInputCropRegion, 4 * sizeof(RpptROI)));
void *d_rowRemapTable, *d_colRemapTable;
if(testCase == 26 || testCase == 79)
{
CHECK_RETURN_STATUS(hipMalloc(&d_rowRemapTable, ioBufferSize * sizeof(Rpp32u)));
CHECK_RETURN_STATUS(hipMalloc(&d_colRemapTable, ioBufferSize * sizeof(Rpp32u)));
CHECK_RETURN_STATUS(hipMemset(d_rowRemapTable, 0, ioBufferSize * sizeof(Rpp32u)));
CHECK_RETURN_STATUS(hipMemset(d_colRemapTable, 0, ioBufferSize * sizeof(Rpp32u)));
}
Rpp32f *cameraMatrix, *distortionCoeffs;
if(testCase == 26)
{
CHECK_RETURN_STATUS(hipHostMalloc(&cameraMatrix, batchSize * 9 * sizeof(Rpp32f)));
CHECK_RETURN_STATUS(hipHostMalloc(&distortionCoeffs, batchSize * 8 * sizeof(Rpp32f)));
}
Rpp32u boxesInEachImage = 3;
Rpp32f *colorBuffer;
RpptRoiLtrb *anchorBoxInfoTensor;
Rpp32u *numOfBoxes;
if(testCase == 32)
{
CHECK_RETURN_STATUS(hipHostMalloc(&colorBuffer, batchSize * boxesInEachImage * sizeof(Rpp32f)));
CHECK_RETURN_STATUS(hipMemset(colorBuffer, 0, batchSize * boxesInEachImage * sizeof(Rpp32f)));
CHECK_RETURN_STATUS(hipHostMalloc(&anchorBoxInfoTensor, batchSize * boxesInEachImage * sizeof(RpptRoiLtrb)));
CHECK_RETURN_STATUS(hipHostMalloc(&numOfBoxes, batchSize * sizeof(Rpp32u)));
}
// create cropRoi and patchRoi in case of crop_and_patch
RpptROI *cropRoi, *patchRoi;
if(testCase == 33)
{
CHECK_RETURN_STATUS(hipHostMalloc(&cropRoi, batchSize * sizeof(RpptROI)));
CHECK_RETURN_STATUS(hipHostMalloc(&patchRoi, batchSize * sizeof(RpptROI)));
}
bool invalidROI = (roiList[0] == 0 && roiList[1] == 0 && roiList[2] == 0 && roiList[3] == 0);
Rpp32f *intensity;
if(testCase == 46)
CHECK_RETURN_STATUS(hipHostMalloc(&intensity, batchSize * sizeof(Rpp32f)));
Rpp32u *kernelSizeTensor;
if(testCase == 6)
CHECK_RETURN_STATUS(hipHostMalloc(&kernelSizeTensor, batchSize * sizeof(Rpp32u)));
RpptChannelOffsets *rgbOffsets;
if(testCase == 35)
CHECK_RETURN_STATUS(hipHostMalloc(&rgbOffsets, batchSize * sizeof(RpptChannelOffsets)));
void *d_interDstPtr;
if(testCase == 5)
CHECK_RETURN_STATUS(hipHostMalloc(&d_interDstPtr, srcDescPtr->strides.nStride * srcDescPtr->n * sizeof(Rpp32f)));
// case-wise RPP API and measure time script for Unit and Performance test
cout << "\nRunning " << func << " " << numRuns << " times (each time with a batch size of " << batchSize << " images) and computing mean statistics...";
for(int iterCount = 0; iterCount < noOfIterations; iterCount++)
{
vector<string>::const_iterator imagesPathStart = imageNamesPath.begin() + (iterCount * batchSize);
vector<string>::const_iterator imagesPathEnd = imagesPathStart + batchSize;
vector<string>::const_iterator imageNamesStart = imageNames.begin() + (iterCount * batchSize);
vector<string>::const_iterator imageNamesEnd = imageNamesStart + batchSize;
vector<string>::const_iterator imagesPathSecondStart = imageNamesPathSecond.begin() + (iterCount * batchSize);
vector<string>::const_iterator imagesPathSecondEnd = imagesPathSecondStart + batchSize;
// Set ROIs for src/dst
set_src_and_dst_roi(imagesPathStart, imagesPathEnd, roiTensorPtrSrc, roiTensorPtrDst, dstImgSizes);
//Read images
if(decoderType == 0)
read_image_batch_turbojpeg(inputu8, srcDescPtr, imagesPathStart);
else
read_image_batch_opencv(inputu8, srcDescPtr, imagesPathStart);
// if the input layout requested is PLN3, convert PKD3 inputs to PLN3 for first and second input batch
if (layoutType == 1)
convert_pkd3_to_pln3(inputu8, srcDescPtr);
if(dualInputCase)
{
if(decoderType == 0)
read_image_batch_turbojpeg(inputu8Second, srcDescPtr, imagesPathSecondStart);
else
read_image_batch_opencv(inputu8Second, srcDescPtr, imagesPathSecondStart);
if (layoutType == 1)
convert_pkd3_to_pln3(inputu8Second, srcDescPtr);
}
// Convert inputs to correponding bit depth specified by user
convert_input_bitdepth(input, input_second, inputu8, inputu8Second, inputBitDepth, ioBufferSize, inputBufferSize, srcDescPtr, dualInputCase, conversionFactor);
//copy decoded inputs to hip buffers
CHECK_RETURN_STATUS(hipMemcpy(d_input, input, inputBufferSize, hipMemcpyHostToDevice));
CHECK_RETURN_STATUS(hipMemcpy(d_output, output, outputBufferSize, hipMemcpyHostToDevice));
if(dualInputCase)
CHECK_RETURN_STATUS(hipMemcpy(d_input_second, input_second, inputBufferSize, hipMemcpyHostToDevice));
int roiHeightList[batchSize], roiWidthList[batchSize];
if(invalidROI)
{
for(int i = 0; i < batchSize ; i++)
{
roiList[0] = 10;
roiList[1] = 10;
roiWidthList[i] = roiTensorPtrSrc[i].xywhROI.roiWidth / 2;
roiHeightList[i] = roiTensorPtrSrc[i].xywhROI.roiHeight / 2;
}
}
else
{
for(int i = 0; i < batchSize ; i++)
{
roiWidthList[i] = roiList[2];
roiHeightList[i] = roiList[3];
}
}
// Uncomment to run test case with an xywhROI override
// roi.xywhROI = {0, 0, 25, 25};
// set_roi_values(&roi, roiTensorPtrSrc, roiTypeSrc, batchSize);
// update_dst_sizes_with_roi(roiTensorPtrSrc, dstImgSizes, roiTypeSrc, batchSize);
// Uncomment to run test case with an ltrbROI override
// roiTypeSrc = RpptRoiType::LTRB;
// roi.ltrbROI = {10, 10, 40, 40};
// set_roi_values(&roi, roiTensorPtrSrc, roiTypeSrc, batchSize);
// update_dst_sizes_with_roi(roiTensorPtrSrc, dstImgSizes, roiTypeSrc, batchSize);
for (int perfRunCount = 0; perfRunCount < numRuns; perfRunCount++)
{
double startWallTime, endWallTime;
switch (testCase)
{
case 0:
{
testCaseName = "brightness";
Rpp32f alpha[batchSize];
Rpp32f beta[batchSize];
for (i = 0; i < batchSize; i++)
{
alpha[i] = 1.75;
beta[i] = 50;
}
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_brightness_gpu(d_input, srcDescPtr, d_output, dstDescPtr, alpha, beta, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 1:
{
testCaseName = "gamma_correction";
Rpp32f gammaVal[batchSize];
for (i = 0; i < batchSize; i++)
gammaVal[i] = 1.9;
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_gamma_correction_gpu(d_input, srcDescPtr, d_output, dstDescPtr, gammaVal, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 2:
{
testCaseName = "blend";
Rpp32f alpha[batchSize];
for (i = 0; i < batchSize; i++)
alpha[i] = 0.4;
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_blend_gpu(d_input, d_input_second, srcDescPtr, d_output, dstDescPtr, alpha, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 4:
{
testCaseName = "contrast";
Rpp32f contrastFactor[batchSize];
Rpp32f contrastCenter[batchSize];
for (i = 0; i < batchSize; i++)
{
contrastFactor[i] = 2.96;
contrastCenter[i] = 128;
}
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_contrast_gpu(d_input, srcDescPtr, d_output, dstDescPtr, contrastFactor, contrastCenter, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 5:
{
testCaseName = "pixelate";
Rpp32f pixelationPercentage = 87.5;
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_pixelate_gpu(d_input, srcDescPtr, d_output, dstDescPtr, d_interDstPtr, pixelationPercentage, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 6:
{
testCaseName = "jitter";
Rpp32u seed = 1255459;
for (i = 0; i < batchSize; i++)
kernelSizeTensor[i] = 5;
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_jitter_gpu(d_input, srcDescPtr, d_output, dstDescPtr, kernelSizeTensor, seed, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 8:
{
testCaseName = "noise";
switch(additionalParam)
{
case 0:
{
Rpp32f noiseProbabilityTensor[batchSize];
Rpp32f saltProbabilityTensor[batchSize];
Rpp32f saltValueTensor[batchSize];
Rpp32f pepperValueTensor[batchSize];
Rpp32u seed = 1255459;
for (i = 0; i < batchSize; i++)
{
noiseProbabilityTensor[i] = 0.1f;
saltProbabilityTensor[i] = 0.5f;
saltValueTensor[i] = 1.0f;
pepperValueTensor[i] = 0.0f;
}
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_salt_and_pepper_noise_gpu(d_input, srcDescPtr, d_output, dstDescPtr, noiseProbabilityTensor, saltProbabilityTensor, saltValueTensor, pepperValueTensor, seed, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 1:
{
Rpp32f meanTensor[batchSize];
Rpp32f stdDevTensor[batchSize];
Rpp32u seed = 1255459;
for (i = 0; i < batchSize; i++)
{
meanTensor[i] = 0.0f;
stdDevTensor[i] = 0.2f;
}
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_gaussian_noise_gpu(d_input, srcDescPtr, d_output, dstDescPtr, meanTensor, stdDevTensor, seed, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 2:
{
Rpp32f shotNoiseFactorTensor[batchSize];
Rpp32u seed = 1255459;
for (i = 0; i < batchSize; i++)
shotNoiseFactorTensor[i] = 80.0f;
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_shot_noise_gpu(d_input, srcDescPtr, d_output, dstDescPtr, shotNoiseFactorTensor, seed, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
default:
{
missingFuncFlag = 1;
break;
}
}
break;
}
case 13:
{
testCaseName = "exposure";
Rpp32f exposureFactor[batchSize];
for (i = 0; i < batchSize; i++)
exposureFactor[i] = 1.4;
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_exposure_gpu(d_input, srcDescPtr, d_output, dstDescPtr, exposureFactor, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 20:
{
testCaseName = "flip";
Rpp32u horizontalFlag[batchSize];
Rpp32u verticalFlag[batchSize];
for (i = 0; i < batchSize; i++)
{
horizontalFlag[i] = 1;
verticalFlag[i] = 0;
}
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_flip_gpu(d_input, srcDescPtr, d_output, dstDescPtr, horizontalFlag, verticalFlag, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 21:
{
testCaseName = "resize";
for (i = 0; i < batchSize; i++)
{
dstImgSizes[i].width = roiTensorPtrDst[i].xywhROI.roiWidth = roiTensorPtrSrc[i].xywhROI.roiWidth / 2;
dstImgSizes[i].height = roiTensorPtrDst[i].xywhROI.roiHeight = roiTensorPtrSrc[i].xywhROI.roiHeight / 2;
}
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_resize_gpu(d_input, srcDescPtr, d_output, dstDescPtr, dstImgSizes, interpolationType, roiTensorPtrDst, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 23:
{
testCaseName = "rotate";
if ((interpolationType != RpptInterpolationType::BILINEAR) && (interpolationType != RpptInterpolationType::NEAREST_NEIGHBOR))
{
missingFuncFlag = 1;
break;
}
Rpp32f angle[batchSize];
for (i = 0; i < batchSize; i++)
angle[i] = 50;
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_rotate_gpu(d_input, srcDescPtr, d_output, dstDescPtr, angle, interpolationType, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 24:
{
testCaseName = "warp_affine";
if ((interpolationType != RpptInterpolationType::BILINEAR) && (interpolationType != RpptInterpolationType::NEAREST_NEIGHBOR))
{
missingFuncFlag = 1;
break;
}
Rpp32f6 affineTensor_f6[batchSize];
Rpp32f *affineTensor = (Rpp32f *)affineTensor_f6;
for (i = 0; i < batchSize; i++)
{
affineTensor_f6[i].data[0] = 1.23;
affineTensor_f6[i].data[1] = 0.5;
affineTensor_f6[i].data[2] = 0;
affineTensor_f6[i].data[3] = -0.8;
affineTensor_f6[i].data[4] = 0.83;
affineTensor_f6[i].data[5] = 0;
}
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_warp_affine_gpu(d_input, srcDescPtr, d_output, dstDescPtr, affineTensor, interpolationType, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 26:
{
testCaseName = "lens_correction";
RpptDesc tableDesc = srcDesc;
RpptDescPtr tableDescPtr = &tableDesc;
init_lens_correction(batchSize, srcDescPtr, cameraMatrix, distortionCoeffs, tableDescPtr);
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_lens_correction_gpu(d_input, srcDescPtr, d_output, dstDescPtr, static_cast<Rpp32f *>(d_rowRemapTable), static_cast<Rpp32f *>(d_colRemapTable), tableDescPtr, cameraMatrix, distortionCoeffs, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 29:
{
testCaseName = "water";
Rpp32f amplX[batchSize];
Rpp32f amplY[batchSize];
Rpp32f freqX[batchSize];
Rpp32f freqY[batchSize];
Rpp32f phaseX[batchSize];
Rpp32f phaseY[batchSize];
for (i = 0; i < batchSize; i++)
{
amplX[i] = 2.0f;
amplY[i] = 5.0f;
freqX[i] = 5.8f;
freqY[i] = 1.2f;
phaseX[i] = 10.0f;
phaseY[i] = 15.0f;
}
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_water_gpu(d_input, srcDescPtr, d_output, dstDescPtr, amplX, amplY, freqX, freqY, phaseX, phaseY, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 30:
{
testCaseName = "non_linear_blend";
Rpp32f stdDev[batchSize];
for (i = 0; i < batchSize; i++)
stdDev[i] = 50.0;
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_non_linear_blend_gpu(d_input, d_input_second, srcDescPtr, d_output, dstDescPtr, stdDev, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 31:
{
testCaseName = "color_cast";
RpptRGB rgbTensor[batchSize];
Rpp32f alphaTensor[batchSize];
for (i = 0; i < batchSize; i++)
{
rgbTensor[i].R = 0;
rgbTensor[i].G = 0;
rgbTensor[i].B = 100;
alphaTensor[i] = 0.5;
}
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_color_cast_gpu(d_input, srcDescPtr, d_output, dstDescPtr, rgbTensor, alphaTensor, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 32:
{
testCaseName = "erase";
init_erase(batchSize, boxesInEachImage, numOfBoxes, anchorBoxInfoTensor, roiTensorPtrSrc, srcDescPtr->c, colorBuffer, inputBitDepth);
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_erase_gpu(d_input, srcDescPtr, d_output, dstDescPtr, anchorBoxInfoTensor, colorBuffer, numOfBoxes, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 33:
{
testCaseName = "crop_and_patch";
for (i = 0; i < batchSize; i++)
{
cropRoi[i].xywhROI.xy.x = patchRoi[i].xywhROI.xy.x = roiList[0];
cropRoi[i].xywhROI.xy.y = patchRoi[i].xywhROI.xy.y = roiList[1];
cropRoi[i].xywhROI.roiWidth = patchRoi[i].xywhROI.roiWidth = roiWidthList[i];
cropRoi[i].xywhROI.roiHeight = patchRoi[i].xywhROI.roiHeight = roiHeightList[i];
}
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_crop_and_patch_gpu(d_input, d_input_second, srcDescPtr, d_output, dstDescPtr, roiTensorPtrSrc, cropRoi, patchRoi, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 34:
{
testCaseName = "lut";
Rpp32f *lutBuffer;
CHECK_RETURN_STATUS(hipHostMalloc(&lutBuffer, 65536 * sizeof(Rpp32f)));
CHECK_RETURN_STATUS(hipMemset(lutBuffer, 0, 65536 * sizeof(Rpp32f)));
Rpp8u *lut8u = reinterpret_cast<Rpp8u *>(lutBuffer);
Rpp16f *lut16f = reinterpret_cast<Rpp16f *>(lutBuffer);
Rpp32f *lut32f = reinterpret_cast<Rpp32f *>(lutBuffer);
Rpp8s *lut8s = reinterpret_cast<Rpp8s *>(lutBuffer);
if (inputBitDepth == 0)
for (j = 0; j < 256; j++)
lut8u[j] = (Rpp8u)(255 - j);
else if (inputBitDepth == 3)
for (j = 0; j < 256; j++)
lut16f[j] = (Rpp16f)((255 - j) * ONE_OVER_255);
else if (inputBitDepth == 4)
for (j = 0; j < 256; j++)
lut32f[j] = (Rpp32f)((255 - j) * ONE_OVER_255);
else if (inputBitDepth == 5)
for (j = 0; j < 256; j++)
lut8s[j] = (Rpp8s)(255 - j - 128);
startWallTime = omp_get_wtime();
if (inputBitDepth == 0)
rppt_lut_gpu(d_input, srcDescPtr, d_output, dstDescPtr, lut8u, roiTensorPtrSrc, roiTypeSrc, handle);
else if (inputBitDepth == 3)
rppt_lut_gpu(d_input, srcDescPtr, d_output, dstDescPtr, lut16f, roiTensorPtrSrc, roiTypeSrc, handle);
else if (inputBitDepth == 4)
rppt_lut_gpu(d_input, srcDescPtr, d_output, dstDescPtr, lut32f, roiTensorPtrSrc, roiTypeSrc, handle);
else if (inputBitDepth == 5)
rppt_lut_gpu(d_input, srcDescPtr, d_output, dstDescPtr, lut8s, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
CHECK_RETURN_STATUS(hipHostFree(lutBuffer));
}
case 35:
{
testCaseName = "glitch";
for (i = 0; i < batchSize; i++)
{
rgbOffsets[i].r.x = 10;
rgbOffsets[i].r.y = 10;
rgbOffsets[i].g.x = 0;
rgbOffsets[i].g.y = 0;
rgbOffsets[i].b.x = 5;
rgbOffsets[i].b.y = 5;
}
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_glitch_gpu(d_input, srcDescPtr, d_output, dstDescPtr, rgbOffsets, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 36:
{
testCaseName = "color_twist";
Rpp32f brightness[batchSize];
Rpp32f contrast[batchSize];
Rpp32f hue[batchSize];
Rpp32f saturation[batchSize];
for (i = 0; i < batchSize; i++)
{
brightness[i] = 1.4;
contrast[i] = 0.0;
hue[i] = 60.0;
saturation[i] = 1.9;
}
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_color_twist_gpu(d_input, srcDescPtr, d_output, dstDescPtr, brightness, contrast, hue, saturation, roiTensorPtrSrc, roiTypeSrc, handle);
else
missingFuncFlag = 1;
break;
}
case 37:
{
testCaseName = "crop";
for (i = 0; i < batchSize; i++)
{
roiTensorPtrDst[i].xywhROI.xy.x = roiList[0];
roiTensorPtrDst[i].xywhROI.xy.y = roiList[1];
dstImgSizes[i].width = roiTensorPtrDst[i].xywhROI.roiWidth = roiWidthList[i];
dstImgSizes[i].height = roiTensorPtrDst[i].xywhROI.roiHeight = roiHeightList[i];
}
startWallTime = omp_get_wtime();
if (inputBitDepth == 0 || inputBitDepth == 1 || inputBitDepth == 2 || inputBitDepth == 5)
rppt_crop_gpu(d_input, srcDescPtr, d_output, dstDescPtr, roiTensorPtrDst, roiTypeSrc, handle);
else
missingFuncFlag = 1;