Evan Harvey

@bartlettroscoe: I don't see this build error showing up in the [PrimaryATDM](https://testing.sandia.gov/cdash/index.php?project=Trilinos&begin=2021-12-08&end=2021-12-14&filtercount=2&showfilters=1&filtercombine=and&field1=buildname&compare1=63&value1=sems-rhel7&field2=groupname&compare2=63&value2=PrimaryATDM) or [SecondaryATDM](https://testing.sandia.gov/cdash/index.php?project=Trilinos&begin=2021-12-08&end=2021-12-14&filtercount=2&showfilters=1&filtercombine=and&field1=buildname&compare1=63&value1=sems-rhel7&field2=groupname&compare2=63&value2=SecondaryATDM) builds. I will raise this github issue during our meeting tomorrow. CC: @jwillenbring, @ZUUL42

@fryeguy52: Would you please look into the module issues on the HPWS machine?

Pending kokkos/kokkos#6124.

@srajama1: Shall I close this as complete?

coo2ccs was not implemented but can be achieved via ccs2crs, coo2crs, crs2ccs.

Closing as complete.

@lucbv: Do you have any notes on this so I can pickup from where you left off or do you want to pair up?

Notes: - Binaries in `issue1663` build directory. - Use https://docs.nvidia.com/compute-sanitizer/ComputeSanitizer/index.html#memcheck-tool - If misalignment originates from a dcomplex view, verify that KOKKOS_ENABLE_COMPLEX_ALIGN is defined in kokkos core.

Relevant snippet from memcheck: ``` ========= Invalid __local__ read of size 16 bytes ========= at 0xdeadbeef in void Kokkos::Impl::cuda_parallel_launch_local_memory(T1 ``` Note that all inverselu invalid reads come from the Blocked...

Note: Cuda/12 wants all addresses 16-byte aligned but, in the BatchedSerialGemm Blocked implementation, we de-reference a address that is 8-byte aligned. TODO: Print out pointer scalar types and their size...