Open tdd11235813 opened 6 years ago
The examples do not do this on desire. This would make the code unnecessarily larger/complex.
All the unit tests as well as the integration tests are already tested for all available accelerators with multiple dimensions and index types.
agreed. I had assumed that the examples shall work on different accelerators. But there should be kind of an advanced example, which show handling of multiple accelerators. If someone uses alpaka, he likely wants to see that.
Isn't this already solved by the <alpaka/example/ExampleDefaultAcc.hpp>
>. See here. And the queue can be created from the accelerator selected by that header.
The alpaka examples should provide all possible compiler switches, for instance:
That can probably a long thing, but can be a code snippet in a header used by all examples?
At the moment
vectorAdd
uses only the CPU backend. For a testing scenario I switched by cmake to CUDA, but I still had to change code inmain.cpp
to test it on CUDA. So what could be a common approach to have a proper code that tests backends as configured by cmake?Edit: fixed code snippet