Looks like someone at Khronos has figured out most of the pieces are there now to do this.
https://www.khronos.org/rfq/openvx-impl ... i-platform
OpenVX on a Pi3B+, yes please
Too bad my skills are still years away from this.
Big kudos to RPF that Pi3B+'s are considered the reference platform for this.
More details in the PDF but start here?
There are forum posts on doe300's VC4CL and I can confirm Arm's Compute Library does compile in Gentoo64.The Khronos OpenVX sample implementation will be available on GitHub at the location below by December 1
Additional suggested resources are:
The AMD Open-Source OpenVX implementation: https://gpuopen.com/compute-product/amd-openvx
The ARM Compute Library: https://developer.arm.com/technologies/compute-library
Open source OpenCL 1.2 project for Raspberry Pi: https://github.com/doe300/VC4CL
A patch to the Khronos sample implementation that leverages the ARM Compute Library to accelerate
many of the OpenVX functions: https://khronos.org/openvx/files/openvx ... ute.tar.gz
To get ARM's Compute library to compile for OpenCL may need massaging the VC4CL lib?
Normally it will use the Mali OpenCL files.
This sort of stuff is why I prefer 64bit OS's and the Khronos target is the 3B+.
This does not look it is intended to run on the single core Pi's.
Multicore ARMv8's are(will be) more common for these applications.
I can imagine Android will be a a major target too as OpenVX will lead to OpenAR/VR?