Simeon Ehrig

Hi, I have a similar problem. I also want to run a `runscript` via `./container.sif`. But in my case I want to ship the container as a demo to other...

CUDA 11.3 does only supports up to GCC 9.x: https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html I had also some problems with GCC 10 and CUDA.

@bernhardmgruber Like you already mentioned yesterday in the VC, our big amount of compiler combinations could be a problem. The `kernel.ompScheduleChunkSize` was only changed in one of the `#if #else`...

@BenjaminW3 What is the time impact of clang-tidy compare to normal compiling?

> I can not quantify clang-tidy compilation time increase. However, we should get most of its advantages by adding a single additional job to our CI, which should not increase...

Most of the changes looks fine for me. In special, I'm really lazy to keep function, class and variable names consistent :sweat_smile: If we apply clang-tidy, we need also document...

> I think we need to discuss how we apply these at some point. Because we had votes on the coding style more than a year ago, but somehow LLAMA...

@tdd11235813 I tried OpenMP with CUDA in a Ubuntu 18.04 container with Clang 9.0 and CUDA 10.0 on a Ubuntu 16.04 hostsystem. Maybe it could solve the problem and specify...

Here is my source-code and the test-environment: source-code: ```C++ #include #include __global__ void foo(float * v, unsigned int start, unsigned int size, float id){ for(unsigned int i = start; i...

No, I didn't fix the header before. I built the clang/llvm from the original Github repo and compiled the source code with the Makefile you see in the previous comment....