iMX93 cross compiling

[matthew_liong]

My goal is to setup TensorFlow for doing ML at iMX93EVK.  Based on imx ML UG, it refers to iMX Yocto project UG. I search nothing useful about cross compiling there. Then I search iMX porting guide, it indicates steps for setting up toolchain.

I tried the suggestion at porting guide, target machine is iMX93EVK, nothing about cortexa53 has been built, instead coretexa9thf can be found.

Is there any suggestion for achieing my goal?  I am a NXP DN veteran, but I am new to iMX stuff. I did careful checking at my input, imx93evk has been correctly accepted and imx93evk related stuffs can be found after building.

[matthew_liong]

Looks like adding eIQ not successful.  Should have a hash value if it is OK?

-------------------------------------------------

Build Configuration:
BB_VERSION = "2.0.0"
BUILD_SYS = "x86_64-linux"
NATIVELSBSTRING = "universal"
TARGET_SYS = "aarch64-poky-linux"
MACHINE = "imx93evk"
DISTRO = "fsl-imx-xwayland"
DISTRO_VERSION = "5.15-kirkstone"
TUNE_FEATURES = "aarch64 armv8a crc crypto"
TARGET_FPU = ""
meta
meta-poky = "HEAD:602922d492351ee747d2ff00f8ed5aab284a706b"
meta-oe
meta-multimedia
meta-python = "HEAD:8f2dc1023482863e2630d1b94052c41ce748b38f"
meta-freescale = "HEAD:06e1cf4099caa53da33c10e1884064f73dc184dd"
meta-freescale-3rdparty = "HEAD:6d8213fc5ec192c33f963b8095d1f01af1574eea"
meta-freescale-distro = "HEAD:fc15f5003043da23212596be7366ae2547c308ad"
meta-bsp
meta-sdk
meta-ml
meta-v2x = "HEAD:c99919c8a838d64e9aa08aa04d8bb2b6fa593d72"

[hector_delgado]

Hi @matthew_liong ,

How was the addition of eIQ not successful? What errors did you receive while creating the image? Or was the image created successfully but there were no signs of eIQ?

Best regards,
Hector.

[hector_delgado]

Hi @matthew_liong ,

I hope you're doing great! If the objective is to have TensorFlow working on the i.MX93 you can review Section 5 of the i.MX Yocto Project User's Guide (nxp.com) that goes into depth on building the image you want. Specifically, you can look at section 5.6.7 NXP eIQ machine learning. It goes into a bit of detail where it also mentions that TensorFlow is included in the imx-image-full image. This is assuming you're running our latest Linux BSP release, if not, let me know what version you are trying to use, and I'll be glad to review it further. Let me know if this is of any help.Have a nice day!

Best regards,
Hector.

[matthew_liong]

Thanks!  Good info.

Yes, ML is one the the goals. But cross-compiling would be another goal before reaching the milestone. Anyway your info can buy me time.

The subject of this thread should fix anyway, and I will leave it opened at this mement.

[hector_delgado]

Hi @matthew_liong ,

I'm glad this was of help. Let me review cross compiling then, and I'll get back to you as soon as possible. Have a nice day!

Best regards,
Hector.