[ROCM] Navi21 Enablement 2: Depthwise kernels (#72682)
Summary:
This PR is a follow up to https://github.com/pytorch/pytorch/pull/69942.
We are adding support to Navi21 GPUs which have a warpsize of 32. We cannot rely on a constant so we have to dynamically look up the warpsize when launching the kernel on the host side. Inside device functions this is not needed and the compiler can correctly detect the correct warpsize to replace the C10_WARP_SIZE constant.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/72682
Reviewed By: soulitzer
Differential Revision: D34174082
Pulled By: ngimel
fbshipit-source-id: 2810e4d1d0f518f20a73a40a0b5c3d71ea120b9e
(cherry picked from commit df003a5bb54cefbf2351f9c1bc8cf91a0609495a)
