Setup Benchmarking Scripts

[SallySoul]

It's early days for this implementation, there's much work to be done.

But,

Things are at a point now that I need to start getting a comparisons with the old implementation
and testing design decisions (#25 , #11 ). Our ground truth for performance is going to on Stampede3, probably on Sapphire Rapids nodes.
I want to start that process.

The old code lives here, we need to be able to run that:
https://github.com/TEAlab/FFTStencils/tree/main

We'll need:

• How do we schedule sapphire rapid cores for non-mpi workload?
• Batch scripts
• Python script for setting up parameter sweep, ect, that's amenable to job submission.
• Python for calling nhls examples
• Python for calling old examples
• Probably some documentation about how to run things.

My initial thought here is to create another repo just for this the Stampede3 experiments.

[SallySoul]

We should be sure to setup Remora, either via python or batch scripts:
https://docs.tacc.utexas.edu/software/remora/

[SallySoul]

Reading Stampede3 User guide, some notes:

There are some build considerations as well, like comparing AVX512 with AVX2
https://docs.tacc.utexas.edu/hpc/stampede3/#building-performance-archflags

Batch scripts for targeting different node types, looks like they're in separate partitions:
https://docs.tacc.utexas.edu/hpc/stampede3/#scripts

Skylake might be a better bet, more of them and they're faster(?)

[SallySoul]

While I procrastinate, I added FFTW3 plan creation strategy and wisdom import / export to the example CLI.
#29

[SallySoul]

Cleaned up initialization a bit, added a rand method to match the 2023 implementation:
#30

[SallySoul]

Expose FFTW3 Measure plan type:
#31