stoa commented on pull request #7742: URL: https://github.com/apache/tvm/pull/7742#issuecomment-823535548
@areusch It would be difficult to use the STM32 without the need to download any package from our site - you need at least the BSP, and the HAL helps a lot. This would be the minimum to be reasonable. We plan to supply the X-Cube with TVM integrated with it. It is easier to generate a project using the X-Cube. Once the project with the board configuration is generated, we can work with it outside of the Cube, if desired. For example, I usually do this - I am traditionally a command line person. The apps/stm32 demo application is a X-Cube generated project (board configuration) where we added our application for Performance Measurement and Validation of ML models. The models are compiled by TVM. I thought it is a good idea to have such a demo bundled with the TVM source base, but you are right - there is still a need to download some basic board support packages. However, notice that this demo application should not add to the CI complexity since it is not part of the CI, just an application note if you wish. If you believe it is better - we can set this application up separately from TVM. Perhaps, there are other applications that are setup in this way ? We would follow their guideline then. -- 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]
