TIOVX User Guide
J721E Linux Performance Report
SoC J721E
Test EVM J721E TI EVM
HOST ARM A72 running Linux at default frequency
Target DSP C66x running Free RTOS at default frequency, L1/L2 Cache Enabled (Cache-only based performance.)




TIOVX Kernel Performance
Index Kernel Variant Frame Size (Pixels) Graph Performance (msec) Node Performance (msec)
1 Accumulate 640x480 (307200) 2.684000 2.591000
2 AccumulateSquare 640x480 (307200) 2.712000 2.631000
3 AccumulateWeighted 640x480 (307200) 1.742000 1.663000
4 Addition U8+U8=U8 640x480 (307200) 2.076000 1.993000
5 Addition U8+U8=S16 640x480 (307200) 3.044000 2.967000
6 Addition U8+S16=S16 640x480 (307200) 4.154000 4.071000
7 Addition S16+S16=S16 640x480 (307200) 3.812000 3.730000
8 Subtraction U8-U8=U8 640x480 (307200) 2.561000 2.477000
9 Subtraction U8-U8=S16 640x480 (307200) 3.093000 3.012000
10 Subtraction S16-U8=S16 640x480 (307200) 3.826000 3.746000
11 Subtraction S16-S16=S16 640x480 (307200) 4.920000 4.834000
12 Not Operation U8 640x480 (307200) 1.589000 1.478000
13 And Operation U8 640x480 (307200) 2.032000 1.952000
14 OR Operation U8 640x480 (307200) 2.009000 1.928000
15 XOR Operation U8 640x480 (307200) 1.999000 1.919000
16 Abs Diff U8 640x480 (307200) 2.538000 2.461000
17 Abs Diff S16 640x480 (307200) 4.837000 4.757000
18 Box 3x3 640x480 (307200) 1.714000 1.636000
19 Dilate 3x3 640x480 (307200) 1.691000 1.610000
20 Erode 3x3 640x480 (307200) 1.681000 1.598000
21 Fast Corners No NMS 640x480 (307200) 13.532000 13.445000
22 Canny 640x480 (307200) 15.079000 14.991000
23 Channel Combine 3 to 1 (RGB) 640x480 (307200) 4.631000 4.549000
24 Channel Combine 4 to 1 (RGBX) 640x480 (307200) 5.966000 5.880000
25 Channel Combine YUYV 640x480 (307200) 2.908000 2.825000
26 Channel Combine 2 to 1 (NV12) 640x480 (307200) 2.386000 2.302000
27 Channel Extract 1 of 3 (RGB) 640x480 (307200) 2.730000 2.614000
28 Channel Extract 1 of 4 (RGBX) 640x480 (307200) 3.948000 3.865000
29 Color Convert RGB=>RGBX 640x480 (307200) 5.315000 5.232000
30 Color Convert RGB=>NV12 640x480 (307200) 3.887000 3.802000
31 Color Convert RGB=>IYUV 640x480 (307200) 3.994000 3.911000
32 Color Convert RGB=>YUV4 640x480 (307200) 4.003000 3.922000
33 Color Convert RGBX=>RGB 640x480 (307200) 5.054000 4.972000
34 Color Convert RGBX=>NV12 640x480 (307200) 4.684000 4.597000
35 Color Convert RGBX=>IYUV 640x480 (307200) 4.737000 4.641000
36 Color Convert RGBX=>YUV4 640x480 (307200) 4.791000 4.710000
37 Color Convert NVXX=>RGB 640x480 (307200) 3.773000 3.692000
38 Color Convert NVXX=>RGBX 640x480 (307200) 4.726000 4.643000
39 Color Convert NVXX=>IYUV 640x480 (307200) 2.379000 2.282000
40 Color Convert NVXX=>YUV4 640x480 (307200) 3.049000 2.968000
41 Color Convert YUVX=>RGB 640x480 (307200) 4.465000 4.381000
42 Color Convert YUVX=>RGBX 640x480 (307200) 5.348000 5.261000
43 Color Convert YUVX=>NV12 640x480 (307200) 3.579000 3.493000
44 Color Convert YUVX=>IYUV 640x480 (307200) 3.327000 3.242000
45 Color Convert IYUV=>RGB 640x480 (307200) 3.943000 3.862000
46 Color Convert IYUV=>RGBX 640x480 (307200) 4.727000 4.639000
47 Color Convert IYUV=>NV12 640x480 (307200) 2.437000 2.357000
48 Color Convert IYUV=>YUV4 640x480 (307200) 3.314000 3.231000
49 Convert Depth U8 to S16 640x480 (307200) 2.423000 2.339000
50 Convert Depth S16 to U8 640x480 (307200) 2.534000 2.452000
51 Convolution 3x3, O: U8 640x480 (307200) 1.882000 1.805000
52 Convolution 3x3, O: S16 640x480 (307200) 2.632000 2.552000
53 Convolution 9x3, O: U8 640x480 (307200) 18.893000 18.810000
54 Convolution 9x3, O: S16 640x480 (307200) 18.339000 18.253000
55 Convolution 3x9, O: U8 640x480 (307200) 24.486000 24.402000
56 Convolution 3x9, O: S16 640x480 (307200) 23.289000 23.195000
57 Convolution 5x5, O: U8 640x480 (307200) 2.235000 2.156000
58 Convolution 5x5, O: S16 640x480 (307200) 2.755000 2.668000
59 Convolution 7x7, O: U8 640x480 (307200) 3.107000 3.029000
60 Convolution 7x7, O: S16 640x480 (307200) 3.300000 3.219000
61 Convolution 9x9, O: U8 640x480 (307200) 37.569000 37.483000
62 Convolution 9x9, O: S16 640x480 (307200) 36.350000 36.263000
63 Equalize Histogram 640x480 (307200) 2.567000 2.487000
64 Gaussian 3x3 640x480 (307200) 1.747000 1.663000
65 Gaussian Pyramid 640x480 (307200) 10.256000 10.159000
66 Half Scale Gaussian Pyramid 640x480 (307200) 1.144000 1.063000
67 Harris Corners 640x480 (307200) 29.504000 29.412000
68 Histogram U8 640x480 (307200) 1.089000 1.006000
69 Integral Image 640x480 (307200) 4.199000 4.117000
70 Laplacian Pyramid U8; Levels = 6 640x480 (307200) 7.713000 7.614000
71 Laplacian Reconstruct U8 640x480 (307200) 10.532000 10.437000
72 LookUpTable U8 640x480 (307200) 3.275000 3.193000
73 LookUpTable U8 640x480 (307200) 1.758000 1.680000
74 Magnitude S16 640x480 (307200) 5.737000 5.654000
75 Mean/Standard Deviation U8 640x480 (307200) 0.911000 0.829000
76 Median 3x3 640x480 (307200) 1.805000 1.722000
77 Min Max Loc U8 640x480 (307200) 1.598000 1.513000
78 Min Max Loc S16 640x480 (307200) 3.001000 2.912000
79 Multiply U8 x U8 = U8 640x480 (307200) 2.088000 2.008000
80 Multiply U8 x U8 = S16 640x480 (307200) 2.043000 1.962000
81 Multiply U8 x S16 = S16 640x480 (307200) 4.306000 4.222000
82 Multiply S16 x S16 = S16 640x480 (307200) 3.824000 3.736000
83 NonLinear Filter mask=5x5; Erosion 640x480 (307200) 2.567000 2.485000
84 NonLinear Filter mask=5x5; Dilation 640x480 (307200) 2.553000 2.468000
85 NonLinear Filter mask=5x5; Median 640x480 (307200) 22.608000 22.510000
86 Optical Flow 5x5 640x480 (307200) 5.325000 5.237000
87 Phase 640x480 (307200) 5.049000 4.966000
88 Remap Bilinear Interpolation *640x480 (307200) 8.932000 8.845000
89 Remap Nearest Neighbor *640x480 (307200) 7.715000 7.629000
90 Scale Image 1 to 2; Nearest Neighbor *640x480 (307200) 1.185000 1.081000
91 Scale Image 1 to 2; Bilinear Interpolation *640x480 (307200) 2.512000 2.429000
92 Sobel 3x3 640x480 (307200) 4.851000 4.765000
93 Threshold Binary 640x480 (307200) 1.656000 1.577000
94 Warp Affine Nearest Neighbor *640x480 (307200) 0.518000 0.436000
95 Warp Affine Bilinear Interpolation *640x480 (307200) 0.847000 0.766000
96 Warp Pespective Nearest Neighbor *256x256 (65536) 0.655000 0.574000
97 Warp Pespective Bilinear Interpolation *256x256 (65536) 1.135000 1.051000

Footnote 1: All optional parameters for relevant kernels are used

Footnote 2: All image sizes are for input parameters unless noted with an asterisk (*) in which case these are the output parameter sizes