Duncan McBain

The easiest thing at this stage is likely to grab it from the VC4CL side - if you look at clCreateProgramWithBinaries, there's a const char ** argument, which will be...

The CMake should pick up the debug version of the ComputeCpp library when building in debug mode, but it's possible that it's not. Can you check the CMake cache (it's...

Yes those are the right libraries. Are you saying that if you change from the *debug* CRT to the *release* CRT, debug builds start working again? If that's the case,...

Hi @j-stephan, thanks for your very comprehensive bug report! Sadly, most built-in functions are not implemented in the ptx backend, so it is the exception rather than the norm that...

Hi @j-stephan, the problem is the way that we hook the user-side SYCL functions (like `cl::sycl::sqrt`) to the actual real device functions. We have it totally implemented for SPIR -...

The PTX backend is missing other bits and pieces, like images. That said, correctly mapping all the builtin functions would go a significant way towards "full" PTX support. From there,...

Hi @j-stephan, I did a little research into the pocl situation last week and it looks like they don't support SPIR on NVIDIA devices at the moment owing to some...

For the first, that shouldn't really change with the maths function used - that is saying that you're asking for the offset of the work-item ID (which is normally zero,...

Thanks @j-stephan, I can confirm the crash happens, though I'm trying to reduce it down as a test case at the moment.

Phew, resurrecting old issues - I have a small repro case and have passed it along to our compiler team. I'll update here with any other information.