在mt8786 （Cortex-A55*6 Cortex-A75*2） 上推理int8模型速度提升不明显

[dreamcatcher-zgr]

detail | 详细描述 | 詳細な説明

loop_count = 64 num_threads = 8 powersave = 0 gpu_device = -1 cooling_down = 0 squeezenet min = 20.73 max = 142.01 avg = 47.55 squeezenet_int8 min = 30.28 max = 114.34 avg = 48.05 mobilenet min = 28.43 max = 100.72 avg = 58.08 mobilenet_int8 min = 27.89 max = 84.14 avg = 48.48 mobilenet_v2 min = 32.89 max = 104.08 avg = 54.17 mobilenet_v3 min = 28.96 max = 91.56 avg = 45.20 shufflenet min = 32.33 max = 121.54 avg = 58.16 shufflenet_v2 min = 27.61 max = 70.47 avg = 44.24 mnasnet min = 33.42 max = 108.99 avg = 55.25 proxylessnasnet min = 41.86 max = 158.48 avg = 60.06 efficientnet_b0 min = 52.60 max = 165.87 avg = 92.01 efficientnetv2_b0 min = 65.24 max = 167.86 avg = 94.25 regnety_400m min = 91.91 max = 298.73 avg = 163.04 blazeface min = 14.94 max = 35.68 avg = 21.15 googlenet min = 84.99 max = 221.07 avg = 127.84 googlenet_int8 min = 82.99 max = 217.91 avg = 145.63 resnet18 min = 45.65 max = 181.71 avg = 92.38 resnet18_int8 min = 76.01 max = 173.99 avg = 105.69 alexnet min = 44.66 max = 109.74 avg = 63.91 vgg16 min = 273.64 max = 519.78 avg = 336.13 vgg16_int8 min = 372.22 max = 631.62 avg = 466.28 resnet50 min = 136.24 max = 300.83 avg = 180.26 resnet50_int8 min = 168.52 max = 372.92 avg = 249.11 squeezenet_ssd min = 77.10 max = 185.13 avg = 112.34 squeezenet_ssd_int8 min = 88.98 max = 198.43 avg = 133.63 mobilenet_ssd min = 56.65 max = 197.05 avg = 102.64 mobilenet_ssd_int8 min = 70.64 max = 166.59 avg = 98.91 mobilenet_yolo min = 194.05 max = 458.28 avg = 299.47 mobilenetv2_yolov3 min = 105.36 max = 205.80 avg = 144.86 yolov4-tiny min = 137.42 max = 298.99 avg = 213.13 nanodet_m min = 79.72 max = 171.35 avg = 111.64 yolo-fastest-1.1 min = 31.20 max = 121.01 avg = 58.47 yolo-fastestv2 min = 37.66 max = 95.55 avg = 57.07 vision_transformer min = 1967.37 max = 2389.48 avg = 2160.50 FastestDet min = 35.38 max = 98.76 avg = 52.36

[dreamcatcher-zgr]

@arnaudbrejeon @pH5 @maxint @baryluk 大佬们来看看呢