It might make sense to pull some of the code out of Node.js and into a faster language like Rust, especially for the compute-heavy calls like AES encryption and WZ parsing (texture decompression). We need to evaluate the performance of the current implementation against a Rust implementation.
- Implement Maple Shanda and AES in Rust
- Add node bindings
- Create a macro benchmark and run both of them with the same inputs (one in Node.js, one in Rust with FFI)
- Report the results
- Do the same for DXT decompression
If the overhead of FFI + running Rust code is far less costly than doing everything in Node.js, we should switch to this method.
Keep in mind this is probably only useful with DXT, because the native crypto module is probably already pretty fast. And it might also be worth it to just install node-dxt, or create another binding for libsquish if I can't get that to work
It might make sense to pull some of the code out of Node.js and into a faster language like Rust, especially for the compute-heavy calls like AES encryption and WZ parsing (texture decompression). We need to evaluate the performance of the current implementation against a Rust implementation.
If the overhead of FFI + running Rust code is far less costly than doing everything in Node.js, we should switch to this method.
Keep in mind this is probably only useful with DXT, because the native
cryptomodule is probably already pretty fast. And it might also be worth it to just installnode-dxt, or create another binding for libsquish if I can't get that to work