ZQPei commented on pull request #9544: URL: https://github.com/apache/tvm/pull/9544#issuecomment-974824269
> Thanks for the PR and I like the solution in general. However, the CUDA scope seems a bit redundant if we already use the CUDA target attribute to control the CUDA arch. I'd suggest, if possible, to completely remove the CUDA scope and fully leverage the target object given by users. > On the other hand, the introduction of `cuda_scope` is probably yet another new interface that would potentially lead to future confusion. I would love to hear about more options if we could possibly make things more concise. Is it possible to detect the "arch" flag in `Target.current()` in the `nvcc.py::compile_cuda`? @comaniac @junrushao1994 Thanks for the suggestions! The initial motivation is to transplant the code of "cuda_target_tarch" and "set_cuda_target_arch" from autotvm to tvm.target, while maintaining the original API as much as possible, this makes me very conservative. The cuda scope is redudant for sure. I will try to remove the cuda scope and take fully usage of `tvm.target.Target` class. -- 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]
