new2f7

Thanks for the comments so far. I'm aware that the duplicated code segments are difficult to maintain. I'm willing to work on this issue but it might take some time...

Sorry for the long delay with this PR, I was busy finalising the associated [paper ](https://dl.acm.org/doi/10.1145/3456669.3456698)and [presentation](https://www.youtube.com/watch?v=hauQGDW6n-8), which are now available. In addition to the requested changes above, I've added...

Thank you very much for your remarks. I look forward to implementing them. > Did you run the tier1 tests with tbb as default already? Does the conformance pass? I'm...

I addressed all previous remarks with the above commits or added a comment why I haven't implemented them (yet). ## question about subdevice support How important is subdevice support for...

> Device interop is important feature/benefit of PoCL. Can you add a smoke test to the test suite which does ping-pong between at least two different device drivers, e.g. tbb+basic...

Hey, I'm sorry, I'm currently not able to work on this PR. Some time ago, I have started some minor work locally which addresses the above comments from late May...

@isuruf Thank you for looking into my stale work. Which two tests of which test suite are failing?

I looked into my notes from 9 months ago and found some unfinished code. I'll just paste it here in case anyone finds it useful. Thanks again for looking into...

Hi, I'm working with @noma on this issue. I tried to use the variable added in https://github.com/pocl/pocl/commit/d6224995936129a2eaef84f2401dd0ef38444500. Unfortunately, I can't see any difference when using `POCL_FORCE_PARALLEL_OUTER_LOOP` or not. I don't...

@Kazhuu Thank you for the detailed description. I had a look at pocl's custom LLVM pass that adds the implicit work-item loop. [WorkitemLoops.cc](https://github.com/pocl/pocl/blob/master/lib/llvmopencl/WorkitemLoops.cc#L546) adds one loop for each dimension (x,y,z)....