Kernels to add

[MrBurmark]

List of kernels to add to the perf suite:

• scan, index list building, Scan Kernels #138

int scan = 0;
for (int i = ibegin; i < iend; ++i) {
  if (x[i] > 0.5) {
    list[scan++] = i;
  }
}
int len = scan;

• Reduction-only kernels. See Add reduction only kernels #114 and Add Tests to Perf Suite #179

• More Atomic kernels. See Add element-wise atomic kernels #115 and Add Tests to Perf Suite #179

• Variants of kernels using RAJA teams to compare against existing kernels that use RAJA::kernel interface

• Variants of kernels using vectorization API when the dust settles on it.

• More kernels from polybench suite that we don't support with RAJA::kernel, but can do with RAJA teams, such as triangular matrix operations.

• Kernels that call functions in different translation units

[rhornung67]

I believe we should be able to do baseline variants of scan kernels for OpenMP and OpenMP target (as of OpenMP 5.0. Is there a reasonable way to do baseline variants of scan kernels for GPU back-ends?

[MrBurmark]

Good call @rhornung67, scan was added in openmp 5.0 so we can conditionally add a baseline for openmp. We can use cub and rocprim for the base variants of cuda and hip scan kernels.