I figure this should probably be split off into its own thread. I was
able to fix the compilation errors with newer versions of OpenCL. See:
https://sourceforge.net/u/kylemaas/enblend/ci/fix-opencl/tree/
However, I am not able to test it yet because it still doesn't like my
ROCM setup for AMD cards. It compiles just fine and doesn't crash when
run, but when trying to execute it, the build process within the self
test phase fails due to being unable to find the ROCM libraries.
Correct way to fix that is probably going to require some more CMake
work, but I wanted to publish this early so you could look over it and
see what all I screwed up. My C++ is a little rusty and I'm not
terribly familiar with OpenCL or Mercurial, but giving it my best
shot. Theoretically, with how small the changes are to port it to the
new unified bindings, it wouldn't surprise me if it works correctly on
Nvidia or Intel cards, but I don't have a good way to test that.
Warm Regards,
Kyle Maas