Daniel Arndt

icon indicating a website link https://www.ornl.gov/staff-profile/daniel-arndt icon indicating an email [email protected]

icon company Oak Ridge National Lab icon location Oak Ridge icon location Large Scale Simulations Computational Scientist. One of the developers of the deal.II finite element library and Kokkos.

Results 728 comments of Daniel Arndt

Deduce workgroup size for SYCL parallel_reduce RangePolicy

`CUDA-9.2-NVCC` failing is unrelated.

Deduce workgroup size for SYCL parallel_reduce RangePolicy

@crtrott I dropped the `mutable` specifier for the lambdas.

Try deducing value type from return type in parallel reduce

Retest this please.

Try deducing value type from return type in parallel reduce

Retest this please.

Try deducing value type from return type in parallel reduce

> Testing lambdas that are generic over the index argument would be nice to add, but I don't see it as necessary. I tried that first but `Cuda` doesn't allow...

Try deducing value type from return type in parallel reduce

This needs more discussion anyway. The main question is if we primarily let the functor or the return value let determine the value type. In the former case, the question...

Try deducing value type from return type in parallel reduce

Again, only `OpenACC` is spuriously failing: ``` Start 33: KokkosCore_IncrementalTest_OPENACC 33: Test command: /var/jenkins/workspace/Kokkos/build/core/unit_test/KokkosCore_IncrementalTest_OPENACC 33: Test timeout computed to be: 1500 33: Error: no GPU available for execution. 33: Raised...

Try deducing value type from return type in parallel reduce

Deducing the value type from the final member function for parallel_reduce without return type takes some more work and I think it's fine to consider doing this in another pull...

Try deducing value type from return type in parallel reduce

> This actually did break compilation of the OpenACC backend, needs fixing. :unamused:

Try deducing value type from return type in parallel reduce

@crtrott I fixed OpenACC.