Work in progress: Parallel Benchmark

[Paramuths]

No description provided.

[willow-ahrens]

Namo, on all the plot labels, could you add a name that reflects:

1. split_rows, split_columns, split_nonzeros
2. static split or dynamic with grain size
3. output memory strategy (scratchspace, atomics, or mutex) if needed

so we would have, e.g., split_rows_static, split_columns_dynamic_50_atomicelement

i think this would help me understand the plots better

[willow-ahrens]

Try thread pinning to see if it improves performance on lanka. https://carstenbauer.github.io/ThreadPinning.jl/stable/examples/ex_pinning_julia_threads/

[willow-ahrens]

Let's also benchmark against taco and suitesparse.