性能数据
大家可以参考测试方法文档对模型进行测试。
ARM测试环境
- 测试模型
- fp32模型
- mobilenet_v1
- mobilenet_v2
- squeezenet_v1.1
- mnasnet
- shufflenet_v1
- int8模型
- mobilenet_v1
- mobilenet_v2
- 测试机器(android ndk ndk-r17c)
- 骁龙855
- xiaomi mi9, snapdragon 855 (enable sdot instruction)
- 4xA76(1@2.84GHz + 3@2.4GHz) + 4xA55@1.78GHz
- 骁龙845
- xiaomi mi8, 845
- 2.8GHz（大四核），1.7GHz（小四核）
- 骁龙835
- xiaomi mix2, snapdragon 835
- 2.45GHz（大四核），1.9GHz（小四核）
- 测试说明
- branch: release/v2.9
- warmup=10, repeats=100，统计平均时间，单位是ms
- 当线程数为1时，
DeviceInfo::Global().SetRunMode设置LITE_POWER_HIGH，否者设置LITE_POWER_NO_BIND
- 模型的输入图像的维度是{1, 3, 224, 224}，输入图像的每一位数值是1
ARM测试数据
fp32模型测试数据
paddlepaddle model
| 骁龙855 |
armv7 |
armv7 |
armv7 |
armv8 |
armv8 |
armv8 |
| threads num |
1 |
2 |
4 |
1 |
2 |
4 |
| mobilenet_v1 |
32.19 |
18.75 |
11.02 |
29.50 |
17.50 |
9.58 |
| mobilenet_v2 |
23.77 |
14.23 |
8.52 |
19.98 |
12.19 |
7.44 |
| shufflenet_v2 |
10.63 |
6.60 |
4.24 |
9.74 |
6.02 |
3.99 |
| squeezenet |
17.44 |
11.39 |
7.50 |
15.33 |
10.04 |
6.91 |
| mnasnet |
20.54 |
12.30 |
7.04 |
17.62 |
10.62 |
6.34 |
| 骁龙845 |
armv7 |
armv7 |
armv7 |
armv8 |
armv8 |
armv8 |
| threads num |
1 |
2 |
4 |
1 |
2 |
4 |
| mobilenet_v1 |
65.28 |
36.37 |
22.88 |
59.27 |
32.62 |
19.57 |
| mobilenet_v2 |
43.40 |
24.33 |
15.43 |
38.15 |
21.77 |
13.81 |
| shufflenet_v2 |
20.09 |
11.55 |
7.57 |
18.45 |
10.91 |
7.16 |
| squeezenet |
32.89 |
21.24 |
13.46 |
30.20 |
19.30 |
12.83 |
| mnasnet |
39.22 |
21.41 |
12.92 |
34.79 |
19.39 |
12.05 |
| 骁龙835 |
armv7 |
armv7 |
armv7 |
armv8 |
armv8 |
armv8 |
| threads num |
1 |
2 |
4 |
1 |
2 |
4 |
| mobilenet_v1 |
92.40 |
51.29 |
33.14 |
86.65 |
48.89 |
27.06 |
| mobilenet_v2 |
63.60 |
36.32 |
23.82 |
61.17 |
33.08 |
19.89 |
| shufflenet_v2 |
27.54 |
16.75 |
11.05 |
24.02 |
14.24 |
8.74 |
| squeezenet |
47.71 |
31.51 |
20.51 |
43.30 |
27.07 |
16.74 |
| mnasnet |
59.17 |
32.38 |
20.71 |
51.29 |
28.32 |
16.95 |
caffe model
| 骁龙855 |
armv7 |
armv7 |
armv7 |
armv8 |
armv8 |
armv8 |
| threads num |
1 |
2 |
4 |
1 |
2 |
4 |
| mobilenet_v1 |
32.14 |
18.70 |
10.91 |
29.49 |
17.48 |
9.60 |
| mobilenet_v2 |
29.87 |
17.64 |
10.68 |
25.82 |
15.58 |
9.29 |
| shufflenet_v1 |
3.96 |
2.80 |
2.05 |
3.64 |
2.70 |
2.04 |
| 骁龙845 |
armv7 |
armv7 |
armv7 |
armv8 |
armv8 |
armv8 |
| threads num |
1 |
2 |
4 |
1 |
2 |
4 |
| mobilenet_v1 |
65.04 |
36.13 |
21.99 |
58.55 |
32.89 |
19.18 |
| mobilenet_v2 |
55.35 |
31.56 |
19.63 |
49.06 |
27.87 |
17.36 |
| shufflenet_v1 |
7.20 |
4.44 |
3.21 |
6.75 |
4.50 |
3.26 |
| 骁龙835 |
armv7 |
armv7 |
armv7 |
armv8 |
armv8 |
armv8 |
| threads num |
1 |
2 |
4 |
1 |
2 |
4 |
| mobilenet_v1 |
94.00 |
52.42 |
31.61 |
85.96 |
45.89 |
49.02 |
| mobilenet_v2 |
81.01 |
46.32 |
29.10 |
81.07 |
43.46 |
42.66 |
| shufflenet_v1 |
10.22 |
6.23 |
4.60 |
10.04 |
6.11 |
4.13 |
int8量化模型测试数据
| 骁龙855 |
armv7 |
armv7 |
armv7 |
armv8 |
armv8 |
armv8 |
| threads num |
1 |
2 |
4 |
1 |
2 |
4 |
| mobilenet_v1 |
19.00 |
10.93 |
5.97 |
13.08 |
7.68 |
3.98 |
| mobilenet_v2 |
17.68 |
10.49 |
5.93 |
12.76 |
7.70 |
4.36 |
| 骁龙845 |
armv7 |
armv7 |
armv7 |
armv8 |
armv8 |
armv8 |
| threads num |
1 |
2 |
4 |
1 |
2 |
4 |
| mobilenet_v1 |
51.37 |
28.11 |
15.50 |
45.06 |
24.47 |
13.80 |
| mobilenet_v2 |
38.90 |
21.64 |
12.33 |
33.03 |
18.71 |
10.77 |
| 骁龙835 |
armv7 |
armv7 |
armv7 |
armv8 |
armv8 |
armv8 |
| threads num |
1 |
2 |
4 |
1 |
2 |
4 |
| mobilenet_v1 |
60.48 |
31.94 |
16.53 |
56.70 |
29.73 |
15.22 |
| mobilenet_v2 |
47.02 |
25.34 |
13.57 |
41.75 |
22.27 |
11.94 |