alpaka

alpaka copied to clipboard

Update documentation copyright

3

The copyright for our documentation is quite outdated. We should update the list of authors and the list of copyright holders. Who should be in these lists? The authors of...

j-stephan

abstract the memory access inside kernels

34

There should not be direct access to memory buffers. This always implies knowledge about the memory layout (row or col) which is not necessarily correct on the underlying accelerator.

BenjaminW3

As discussed this morning, a test for this functionality is available at https://github.com/fwyzard/test-rocm-cuda-alpaka . With the current `develop` branch (actually with #1678 to make the explicit types available), the test...

bernhardmgruber

[RFC] Use SPDX license identifiers

2

All alpaka headers use a license statement at the top. This adds a lot of noise and there is a more compact way to specify the license: SPDX license identifiers....

bernhardmgruber

`QueueGenericThreadsNonBlocking` technically illegal with OpenMP

3

I ran into an issue with the new `FenceTest` which occasionally crashes after succeeding with NVHPC for any OpenMP backend. This test runs kernels in parallel using separate non-blocking queues....

jkelling

clang-14 was released a few days ago. We should add it to the CI.

bernhardmgruber

Omp2Blocks aborts when `num_threads = 1` while `omp_get_max_threads > 1`

7

https://github.com/alpaka-group/alpaka/blob/a2315638266b9c2a429d34e4e8e695405045d486/include/alpaka/kernel/TaskKernelCpuOmp2Blocks.hpp#L879 This check fails because there is [no parallel region created](https://github.com/alpaka-group/alpaka/runs/4853180884?check_suite_focus=true#step:5:5592) after going through [`omp parallel`](https://github.com/alpaka-group/alpaka/runs/4853180884?check_suite_focus=true#step:5:5590). This happens in the CI with Clang 11, but not with GCC 11. I...

jkelling

Improve CI efficiency

14

Discussed offline with @bernhardmgruber: The current way of running the CI every time when a PR is merged is very inefficient. We just tested the code base against the PR,...

j-stephan

As we discussed today, triggered by how the hello world example looks in new readthedocs #1093 , there are a couple of things we can do to improve readability and...

sbastrakov

The Doxygen API is hard to read because they provide information about the users and the internal API. Making the Doxygen API easy to read without losing information is not...

SimeonEhrig