I’m currently trying to reduce the development back and fourth and continuous integration build pipeline for an application we are developing using a custom driver driver built for two orins (one in PCIe target mode one operating as a PCIe root complex). We are currently hoping to use the driver to pass messages from various sensors for an extremely light weight embedded edge application. However, there is a big strain on hardware for us to make meaningful driver development headway, especially in the way of performance, because it takes a lot of time to compile, build, deploy, then test the small incremental changes that are being made to the driver. Normally, in a situation like this, I would build a user mode linux kernel and corresponding driver and then mock the PCIE hardware with a separate process. However, in this case, there are really two different nvidia Orin units, one in target mode utilizing the target mode driver, and another NVIDIA orin working as the root complex. I’m curious if any one out there has ever utilized a user mode Linux kernel working as a PCIE endpoint to better test the inter workings of two different orins, one working in target mode and the other working as the root complex. It seems like a necessity for developing a PCIe application where your PCIe card has a full running operating system.
Is there a way to test drivers for an Nvidia orin operating as an endpoint without 2 orins (preferably using something like user mode linux)?
1 Like
Sorry. Don’t know the answer.
Also please note this is forum for CUDA-GDB, the develop tools. You may want to try asking in Orin forum directly.
1 Like
This topic was automatically closed 2 days after the last reply. New replies are no longer allowed.