mikolalysenko/ndarray-experiments

Benchmark measurements

mscdex opened this issue · 1 comments

mscdex commented

I have some questions about the benchmarks here:

  1. Are you intentionally benchmarking the creation of arrays also or are you really wanting to just benchmark operations on arrays?
  2. Shouldn't the results from the benchmarks be checked against some expected values to verify the implementation is performing the correct operation(s)?

With regards to 1, I have made changes locally to just time the array operations and not the creation, since operations are likely to happen more frequently than creation. I have also made some other adjustments to the output to make sure everything lines up correctly and is more readable. I can submit PRs for these if you like.

mikolalysenko commented

Thanks! If you have a pull request I'll add it in.

Regarding point #2, it would probably be interesting to separate out creation time from the the benchmark itself. If you want to take a crack at it, I'll also accept a pull request.