Daniel Arndt

Results 789 comments of Daniel Arndt

With the last commit, the output is now ```bash > ./variorum-print-verbose-energy-example _INTEL_GPU_ENERGY_USAGE Host: x1921c0s0b0n0, Socket: 0, DeviceID: 0, Energy: 0.000000 J _INTEL_GPU_ENERGY_USAGE Host: x1921c0s0b0n0, Socket: 0, DeviceID: 1, Energy: 0.000000...

Currently, I'm storing the initial energy in a global static variable. I noticed that `initAPMIDG` and `shutdownAPMIDG` are called for every API call and not in some function that would...

> It might be better to just do static memory allocations here as opposed to `malloc/free`. It is not a lot of additional memory (only 3 or 4 variables and...

> What we can do is try to have the `SequentialHostInit` overwrite the `WithoutInitializing` sacrificing some performance Yes, that's what I would do.

> I see this message in the output of the failing Stratimikos and Panzer tests: That looks like issues with `cudaMallocAsync`, see https://kokkos.org/kokkos-core-wiki/known-issues.html?highlight=known+issues#cuda, and https://kokkosteam.slack.com/archives/C5BGU5NDQ/p1726216998539829.

We decided to make `-DKokkos_ENABLE_IMPL_CUDA_MALLOC_ASYNC=OFF` the default again in `Kokkos` (but that will likely only be visible in `Trilinos` with the next release).

We have seen ``` /builds/kokkos/kokkos/core/src/Cuda/Kokkos_Cuda_Parallel_Range.hpp:44:16: required from 'class Kokkos::Impl::ParallelFor' /tmp/tmpxft_00001a05_00000000-6_TestCuda_GraphAtomicLocks.cudafe1.stub.c:28:547: required from here /builds/kokkos/kokkos/core/src/traits/Kokkos_ExecutionSpaceTrait.hpp:46:8: error: 'Kokkos::Impl::ExecutionSpaceTrait::mixin_matching_trait' declared with greater visibility than its base 'Kokkos::Impl::AnalyzeExecPolicy' [-Werror=attributes] 46 | struct mixin_matching_trait :...

How would you initialize a default-constructed object instead?

> If the default constructed pool is not actually useful, maybe we shouldn't allow that either Yes, but non-default-constructible classes are a pain as well since it makes it much...

> I didn't modify `kokkos_arch.cmake` so I don't understand why the format check is failing See https://github.com/kokkos/kokkos/pull/7945.