zhiics commented on pull request #6076: URL: https://github.com/apache/incubator-tvm/pull/6076#issuecomment-659865400
The way you currently do is more like a hack to make the example work. We need to more systematically consider different instructions/ir nodes. ShapeFuncs are always on CPU and kernels are on GPU. But there are also more than that. For example, we will introduce some instructions in IR to simplify VM runtime, i.e. shapeof, etc. These instructions all need to be considered. We are currently working on some higher priority tasks. After that, I will work on upstreaming the code. So Let's close this for now. ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected]
