Post
at least 90% of the improvement comes from architecture and focus since I am starting from scratch I don't have 14 years of baggage and backward compatibility to deal with my approach for performance is similar to the "Page Load Test" (PLT) benchmark the Safari team used, see bench:mermaid in Pele
0 reposts4 likes
my experiments made me confident I could get to 6x smaller and 500x faster, so that's my budget Mermaid has 33 types and endless uses, any perf cost I add, I have to find a perf improvement it's death by a thousand cuts, so MI is helpful for finding optimization space
might I offer github.com/Saeris/deopt... to help with discovering more perf optimization opportunities? I've had some mild success using it to find areas of improvement in a variety of projects
Some more info on the PLT here (2019), scroll down to On Creating Automated Process Tooling: blog.jim-nielsen.com/2019/book-no...