bash run.sh
using m=10000000 n=64
CUDA kernel takes 10 ms
verified
Input shape torch.Size([10000000, 64])
TORCH max takes 24.503946 ms
Tested on sm70 and sm86(RTX 4060Ti)
bash run.sh
using m=10000000 n=64
CUDA kernel takes 10 ms
verified
Input shape torch.Size([10000000, 64])
TORCH max takes 24.503946 ms
Tested on sm70 and sm86(RTX 4060Ti)