- Windows 10
- CUDA 11.1
- TensorRT 8.0.1.6
- Cudnn 8.2.1
- Cudnn Convolution Algorithm execution time Comparison (1000 iteration)
- input [1,3,224,224] weight [32,3,3,3] bias [32]
| IMPLICIT_GEMM | IMPLICIT_PRECOMP_GEMM | GEMM | FFT | FFT_TILING | WINOGRAD | WINOGRAD_NONFUSED | |
| Avg Duration time [ms] | 1.612 ms | 0.110 ms | 0.135 ms | 1.655 ms | 1.305 ms | 0.105 ms | 2.868 ms |