Neural Network Inference on Pixel 4
17.5Baseline Latency (ms)Fine-Grained CPU-GPU Co-Execution
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Fine-Grained CPU-GPU Co-ExecutionNeural Network=ResNet-18, CPU Threads=32025.10 | 17.5 | 11 | 1.59 | 11.4 | 1.54 | |
| Fine-Grained CPU-GPU Co-ExecutionNeural Network=ResNet-34, CPU Threads=32025.10 | 37.5 | 22.2 | 1.69 | 22.5 | 1.67 | |
| Fine-Grained CPU-GPU Co-ExecutionNeural Network=VGG16, CPU Threads=32025.10 | 83.3 | 70.8 | 1.18 | 73 | 1.14 | |
| Fine-Grained CPU-GPU Co-ExecutionNeural Network=Inception-v3, CPU Threads=32025.10 | 99.5 | 59.1 | 1.68 | 61.5 | 1.62 |