Daniel Arndt

Results 1005 comments of Daniel Arndt

Can you try ```terminal git rebase upstream/master git rebase -i upstream/master ``` to only squash your commits in the second step?

Judging from the merge commit messages, it seeem your remotes name is `dealii` instead of `upstream`. Since you have merge commits, it might be easiest to just do ``` git...

> It does not work. I am thinking it can be faster to open a new PR with one clean commit. I don't think it makes much of a difference...

> How do I push this changes out? Just like you did before but add `-f` to force the push.

``` Kokkos_HIP_ReduceScan.hpp:382:36: error: comparison of integers of different signs: 'size_t' (aka 'unsigned long') and 'const short' [-Werror,-Wsign-compare] [HIP-ROCm-6.2-CXX20] 382 | for (size_t i = threadIdx.y; i < word_count.value; i +=...

Note that most of these functions are provided in the Intel Math Function Device Libraray (https://www.intel.com/content/www/us/en/docs/dpcpp-cpp-compiler/developer-guide-reference/2025-2/imf-transcendental-math-functions.html), e.g., as `sycl::ext::intel::math::lround`.

What about the math functions for half precision and quarter precision types?

SYCL CI is failing with ``` /home/runner/_work/kokkos/kokkos/kokkos/core/src/impl/Kokkos_Half_MathematicalFunctions.hpp:215:26: error: no member named 'impl_lround' in namespace 'Kokkos::Impl' 215 | return Kokkos::Impl::impl_lround(x); | ~~~~~~~~~~~~~~^ /home/runner/_work/kokkos/kokkos/kokkos/core/src/impl/Kokkos_Half_MathematicalFunctions.hpp:218:26: error: no member named 'impl_llround' in namespace 'Kokkos::Impl'...

> Maybe that is a good idea. @ndellingwood @masterleinad thoughts? I think it wasn't quite clear if any SNL machine is ready for testing with Cuda 13. I find GitLab...

I'm fine with just adding a build on the HPSF site as well.