I would love to follow up and add a preprocessing stage to the TVM target creation. It is useful for users/developers to clearly know the complete target info at the earliest stage.
For example, the LLVM NVPTX backend adds `-mtriple=nvptx64-nvidia-cuda`, and `-mcpu=sm_xx` if missing. We want to move those information at the beginning stage of target creation to make sure the information is clearly presented to developers. --- [Visit Topic](https://discuss.tvm.ai/t/rfc-tvm-target-specification/6844/54) to respond. You are receiving this because you enabled mailing list mode. To unsubscribe from these emails, [click here](https://discuss.tvm.ai/email/unsubscribe/1ebf55fb8bd81f27b3340b149cca5ab35f13c947846d6fe81642e582d58bdd35).
