Hi, I've got a c++ implementation of RaptorQ. It's a part of another project and private now, but it will be released eventually.

First, great work. You have achieved a really nice performance on 10K bytes.

But I have concerns about performance with bigger symbol_count.
For 10K data and symbol_size=512, symbol_count should be 20.

I've tried to change elements to 512 * 50000 in you encode 10KB benchmark, and it doesn't seem to finish in a reasonable amount of time.

Am I doing something wrong?
I'm running it with cargo bench --features benchmarking.
Have you tested that it did make the codec to scale linearly to larger blocks?

As I recall rfc is quite cryptic about details critical for performance.