Building OpenCV 3.4.6 for Xavier - is VFPV3 supported?

Autonomous Machines Jetson & Embedded Systems Jetson AGX Xavier
#1

Hi,

I have KCF tracker code whcih heavily relies on CPU rather than GPU, but I I would like to test its performance on Xavier.

When building OpenCV 3.4.6, I try turn on the NEON & VFPV3 flags
-D ENABLE_NEON=ON
-D ENABLE_VFPV3=ON
But CMAKE fails on VFPV3. The NEON flags by itself passes CMAKE.

Other guides (e.g. buildOpenCVXavier/buildOpenCV.sh at master · jetsonhacks/buildOpenCVXavier · GitHub) do not include these flags when building OpenCV.

Per the public info, Carmel is a ARMv8.2 comliant vs the A53 which is ARMv8.0. I did find some guidance for the ARM A53 case that VFPV3 is supported:

Is VFPV3 supported on Carmel architecture.

BR,
Amit

OpenCV, CUDA, Python with Jetson Nano
#2

Some similar topic for your reference.

https://devtalk.nvidia.com/default/topic/1038699

-D ENABLE_VFPV3=ON
But CMAKE fails on VFPV3. The NEON flags by itself passes CMAKE.

What error did cmake return?

#3

Thanks.
I checked thw post you suggested. I dont think its relevant since I need to set compiler flags for the entire OpenCV package and not my own application.

CMAKE returns the following error:

CMake Error at cmake/OpenCVCompilerOptimizations.cmake:477 (message):
Required baseline optimization is not supported: VFPV3
(CPU_BASELINE_REQUIRE=;VFPV3;NEON)
Call Stack (most recent call first):
cmake/OpenCVCompilerOptions.cmake:277 (include)
CMakeLists.txt:713 (include)

#4

I’m having the same problem, how should cmake command be modified?