Open sethaxen opened 2 years ago
The PRs contributing backends have used a simple benchmark to decide which overloads are needed. I suggest we formalize this to a benchmark we can run to assess performance impact for any backends of any PRs changing the fallback implementations.
The PRs contributing backends have used a simple benchmark to decide which overloads are needed. I suggest we formalize this to a benchmark we can run to assess performance impact for any backends of any PRs changing the fallback implementations.