ArmPL TPL support for KokkosBlas and KokkosBatched

[e10harvey]

• Check for ArmPL version 21 TPL installations
• Add CMake TPL support. (@e10harvey) - cmake: Add ARMPL TPL support #880
• Add KokkosBlas TPL support: gemm, iamax, scal, copy (Kokkos copy). (@vqd8a) - cmake: Add ARMPL TPL support #880
  • Collect performance in Adelus
    • gemm
    • iamax
    • scal
    • copy
• Add KokkosBatched TPL support to cover routines used by https://github.com/kokkos/kokkos-kernels/blob/master/perf_test/batched/KokkosBatched_Test_BlockTridiagDirect.cpp. (@e10harvey, @vqd8a)
  • LU (@vqd8a)
  • TRSM (@vqd8a)
  • GEMM (@e10harvey) - src/batched: Add armpl dgemm support #1256
  • TRSV (@e10harvey)
  • GEMV (@e10harvey)
• Collect BlockTridiagDirect performance in KokkosKernels

[vqd8a]

@e10harvey Thanks for adding the CMake for ARMPL. I would like to add two comments:

1. For ARMPL's BLAS, could you please also enable KOKKOSKERNELS_ENABLE_TPL_BLAS in the KokkosKernels_config.h when KOKKOSKERNELS_ENABLE_TPL_ARMPL is defined, so that we can use the current BLAS TPL support in Kokkos Kernels?
2. It looks to me that single-threaded ARMPL (libarmpl.so) can be found

 =======================
KokkosKernels ETI Types
   Devices:  <OpenMP,HostSpace>
   Scalars:  double
   Ordinals: int
   Offsets:  int;size_t
   Layouts:  LayoutLeft

KokkosKernels TPLs
   ARMPL:       /opt/arm/armpl-20.3.0_A64FX_RHEL-8_gcc_aarch64-linux/lib/libamath.so;/opt/arm/armpl-20.3.0_A64FX_RHEL-8_gcc_aarch64-linux/lib/libarmpl.so
=======================

Can we also find multi-threaded ARMPL (libarmpl_mp.so) when OpenMP is enabled?

[e10harvey]

1. For ARMPL's BLAS, could you please also enable `KOKKOSKERNELS_ENABLE_TPL_BLAS` in the `KokkosKernels_config.h` when `KOKKOSKERNELS_ENABLE_TPL_ARMPL` is defined, so that we can use the current BLAS TPL support in Kokkos Kernels?

Yes.

Can we also find multi-threaded ARMPL (libarmpl_mp.so) when OpenMP is enabled?

Yes.

I will flag you in the PR for this.

[vqd8a]

@e10harvey Thanks, Evan.