vinx13 commented on PR #16302: URL: https://github.com/apache/tvm/pull/16302#issuecomment-1873545551
This may also change behavior of compile other cuda sources other than `cutlass_fpA_intB_gemm` kernels. What about doing the following: * Remove the hardcoded architectures in NVCC flags in `cutlass_fpA_intB_gemm` so that it depends on the new cmake behavior of setting `CMAKE_CUDA_ARCHITECTURES` * Provide a default value `native` if `CMAKE_CUDA_ARCHITECTURES` is not defined (which is generally larger than 52 for not-too-old gpus) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
