DeepStream App build system

• Hardware Platform (Jetson / Nano and Xavier)
**• DeepStream Version 5.1 **
• Issue Type( questions,)

what are people using as a build machine for building and deploying DeepStream applications.

I’ve built a single Nano using the SDK manager that i’m using as a development machine.
For reproducible and automatic builds for other systems we have specific build machines.

Is it possible to build the binary of the DeepStream app, in for example a docker image running on x86 ?
or is it common practise to just use a dedicated Nano/Xavier to do the building, then wrap that into a docker image for redistribution to all the edge devices.

Just wondering what ya’ll do.

If anyone is doing cross compilation for the nano/xavier… which toolchain should i use to match the one that the devtools on the device have from :

found one that works (gcc-linaro-7.3.1-2018.05-x86_64_aarch64-linux-gnu.tar.xz)
grabbed that one for the toolchain and imaged my development Nano sd card to a file, mount it and point to it with
–sysroot= to the CFLAGS
some more pointing of the include directories to the sysroot -I for the compile and -Wl,-rpath-link=$(TARGET_ROOTFS)/lib/aarch64-linux-gnu (plus some more ) for the linking.

all good.

so it looks like using the cross compile may be a valid option, then overlaying the output of the cross-compile to the base docker image


1 Like