Favorite part is how he mixes the register allocator with a chain of seds.<p>I did something similar with a youtube downloader I wrote, using a long chain of seds. It is not as beautiful as Python but it is smaller and faster.<p>When I shared it with HN the youtube-dl author called it "unmaintainable". By who? I have had no problems maintaining it. :)
This looks like it might be interesting, but the most recent date listed is 2006, with little to suggest that anything has been done since then. Does anyone know if this is nothing more than an historical artifact?
Has anyone ever really understood how it happens that code produced with qhasm is so fast? I understand code unrolling, but DJB mostly gets speed because he understands the sequence of instructions in an algorithm well. As far as I remember you had to have a machine architecture spec for this tool. So I wonder if it's a more optimized register allocator?