Pekka Jääskeläinen

FWIW, run_cuda_tests works for me on my desktop NVIDIA GPU. I use it for final pull request testing before I push anything to master. I haven't tried on Jetson TX2...

Unfortunately me nor others in the CPC group currently have the time for in-depth debugging sessions of pocl issues that do not affect our research tracks (read: publication results). However,...

@isuruf are you planning to look at this bug before the pocl 1.6 release? Shall we wait?

OK, I assume this is not a regression.

Please retry if you can still reproduce this with the latest master revision and/or the latest LLVM.

Yeah, I just noticed the same thing while testing your PR. For now, I changed all of them to lower case for consistency (and to match what's written in the...

Since fixed in LLVM release branch, not urgent for 1.6.

Does pocl work at all for you? That error message is usually correct, meaning that your kernel builtin lib is corrupted (e.g. refers to missing symbols). POCL_DEBUG=1 might help too.

I did it to make the run_cuda_tests suite to pass reliably with my pull request test PC. Otherwise, it failed unreliably. Now it has passed for a while without crashing....

Please note that we are nearing the release and since it's so late already, this cannot hold it back.