Sounds like a search problem. I wonder why sharing knowledge works so well when using Stack Overflow but not for his research results?

Because the purpose of research is to produce papers, not disseminate information or help others.

See how many CS papers talk about various algorithms without also providing the source code of said implementation.

OK, but for any individual researcher (such as the original poster's friend) what if they want to try harder?

Add a github repo with a permissible license, write a blog post, add a stack overflow answer to some relevant questions.

To be fair there aren’t many questions on stack overflow about, “How can I design my hyperblock selection algorithm for improved performance on a block-structured data flow processor?”

There might be if you first go out of your way to answer questions about dataflow throughput by pointing people toward block-structuring and mention keywords like "hyperblocks" there. People already do this; I've noticed a good few times now that I've done a number of increasingly-specific searches as I understand a problem better, and each time the answer that actually guides me to the next step comes from the same person/place.