Native gcc compiler for Apalis iMX8qm

Dears,

Would you let me have some idea on aarch64-tdx-linux-gcc native compiler running on Apalis iMX8QM SOM ? What can I make it with dunfell-5.x.y yocto branch?

Greetings @toddhwang,

May I ask why you would want to natively compile rather than cross-compile? For almost every situation natively compiling will be slower than cross-compiling on a sufficiently powerful desktop.

Best Regards,
Jeremias

Hi Jeremias.

Appreciated your reply along with reasonable interests. Let me share with you why moving to performance-threatening jobs on top of embedded board. Actually I am not too much sure that 2 x superscalar CortexA72 cores will get me the result of game changer in embedeed application. My application is some of intelligence stuff and it requires to run “OpenCV” Python stuff, and I know that I could enjoy the mercy of software readiness of Yocto distribution from the Toradex. I can successfully and easily download opencv and imutils by using "pip3 install " command but the problem happens on learning. The learning process relies on “scikit-learn” python component and it tries to keep running aarch64-tdx-linux-gcc compiler insuide of the board.

#1. I am very novice to Python stuff. Can I add “imutils” “opencv-python” “scikit-learn” as packages in a recipe ? Can you provide the files for those for me?

#2. Otherwise, how to make native compiler running on the Apalis iMX8qm?

Thanks,
Todd.

@jeremias.tx , single bitbake launch takes on average PC about minute just to tell that it has nothing to do. Linux host updates seem sometimes launch a tone of remakes and rebuilds. Some libraries are very not cross compiling friendly. Perhaps for Yocto guru’s it is not a problem, but for me it certainly is.

@toddhwang, I’m not sure about iMX8qm. For iMX7 I add to IMAGE_INSTALL += string in your image bb:

    \
    packagegroup-sdk-target wget git \
    m4 libtool patch tar automake libffi-dev \

This makes bitbake adding to the image GCC and other basic tools.

Best Regards,
Edward

Regarding the python packages you can search for the yocto recipe here: https://layers.openembedded.org/layerindex/branch/master/layers/

Usually it’s fairly up-to-date. Worst case the recipe doesn’t readily exist and you may need to make your own.

As for the native compiler please see the above answer from Edward.