Ronan Keryell

That said, Ubuntu 18.04 and GCC 7 are quite old now, so at least you should try with Ubuntu 20.04.

Yes, the AMD ROCm DevOps story is suboptimal... :-( But we can hope for some progress, even if it slow to happen.

Yes, this is the annoying part because I was using this code before without `this` and I had some warnings with C++20. So, I added it to avoid warnings with...

Thank you for the contribution. I think it would be nice to integrate this somehow into https://github.com/triSYCL/triSYCL/blob/master/doc/testing.rst and https://github.com/triSYCL/triSYCL/blob/master/doc/cmake.rst By the way we do not use `.md` files in this...

@dhollman : is your presentation publicly accessible somewhere?

Thank you for the slides! Nice there is also some implementation of this.

@chriskohlhoff In SYCL we focus on heterogeneous computing, where we do not have a uniform memory model following the standard C++ one. So typically you want some buffers to be...

Crazy! At least you have found the problem, so it should be easier to fix it now. Is it a side effect of all the `alignas(sizeof(DataType) * 4)` and so...

You can look at the LLVM IR bitcode and look at the type definition of the appropriate `class.vec...` to understand what Clang/LLVM think it is.

A simple way to start is to add a unit test in triSYCL testing for some size & alignment properties. Be careful that the case with 3 element might be...