Dynamic memory allocations were happening each time a recording_function was called. Additionally having a lambda function to handle conversions is only needed when the type of the thing being recorded doesn't match the recorder type. If the types match then we can simply save a pointer and get much better performance.
Should be quite a bit faster based on results from a short Buzzard scenario
Before:
After:
Dynamic memory allocations were happening each time a recording_function was called. Additionally having a lambda function to handle conversions is only needed when the type of the thing being recorded doesn't match the recorder type. If the types match then we can simply save a pointer and get much better performance.
Should be quite a bit faster based on results from a short Buzzard scenario Before: After: