Arturo Vargas

Hi @delcmo , `RAJA_DEVICE` is just a macro for CUDA/HIP's `__device__` qualifiers. In general, if you want to call a function on the device you need to add the `__device__`...

> Hi @artv3 thanks for your reply. This is what I am after. I just started with RAJA and may have overlooked a few things in the documentation. > >...

Sounds good, I can definitely help. I use RAJA in a couple of applications and can help with integration/usage questions.

Hi @delcmo, Regarding 1. The `compute_macro_srt2` function signature expects a double * for f, g, not `RAJA::View ` types which is what led to the compilation error. If you would...

Hi @delcmo, the RAJA Views just hold a pointer, the host code is still responsible for allocating the memory. Views will work on both the **host and the device** The...

Hi @delcmo, I took a look at your code, and one possible implementation with RAJA Teams which enables running sequentially or with CUDA could look like this: This first part...

One additional detail to note is that I am using macros to guard the CUDA code, this is needed in case we build on a CPU only platform where the...

Hi @delcmo , yes those are specific to Teams. There is partial documentation in the develop branch of read the docs: https://raja.readthedocs.io/en/develop/sphinx/user_guide/feature/loop_basic.html#team-based-loops-raja-launch I think we currently point our release the...

> @artv3 > > in the post where you converted the code from C++ to RAJA, you left the loop over the direction unchanged. > > ``` > for (int...

> Hi @artv3, > > if I understand you correctly, I can keep the C++ for loop syntax as it is. Does that mean each GPU thread will perform a...