Where is torch.matmul implemented, especially the part that runs on the GPU?
The whole project is 2M lines of code. I tried to grep the sources of the 1.8.2 release, but have trouble finding this function. I'm guessing it's generated from something...
It should be implemented below: https://github.com/pytorch/pytorch/blob/master/aten/src/ATen/native/LinearAlgebra.cpp#L1450
The matmul function included by the Functions.h and ATen.h included Functions.h
And inside pytorch api(pytorch/torch/csrc/api/include/torch/linalg.h), it includes <ATen/ATen.h>
So:
pytorch/torch/csrc/api/include/torch/linalg.h > ATen > Functions > LinearAlgebra > matmul
