The current code looks up pinyin word by word. We need to investigate whether multithreading brings any performance benefits, considering the additional overhead from context switching. Right now, enabling multithreading introduces extra compiler options /MT with msvc, that may propagate throughout the entire project.
If we wrap multithreading around the inference, and each inference also includes multithreading for pinyin processing, would the overhead of having multithreading within multithreading be too large?
The current code looks up pinyin word by word. We need to investigate whether multithreading brings any performance benefits, considering the additional overhead from context switching. Right now, enabling multithreading introduces extra compiler options
/MTwith msvc, that may propagate throughout the entire project.If we wrap multithreading around the inference, and each inference also includes multithreading for pinyin processing, would the overhead of having multithreading within multithreading be too large?