hermes

Say your code has two strings "split pea" and "pea soup". In the binary file, you can overlap the suffix of the first with the prefix of the second to form "split pea soup": the "pea" is shared, and this reduces both memory usage and binary size. "split pea soup" is called a superstring of "split pea" and "pea soup" because it contains them both. An optimal superstring packs strings as closely as possible, and its construction is a moderately well-studied problem. This is the best paper I found: https://www.sciencedirect.com/science/article/pii/0304397588... After implementation I found that "optimal superstrings" compress poorly because the string ranges are unpredictable, and I later did some work to make the string ranges compress better even if the superstring gets longer. The Hermes superstring optimization was time unwisely spent bu

No one’s looking at Static Hermes? https://github.com/facebook/hermes/commits/static_h

It's "Quick" in the sense that it's a bytecode interpreter instead of a heavyweight JIT, so much faster startup times (although overall compute performance is much slower). And the library is much smaller than more heavyweight engines like V8, so good for reducing overall app size. Recently Facebook had to implement a whole new non-JIT Javascript engine just for the purpose of optimizing startup times for React Native apps (https://github.com/facebook/hermes). If you consider how even Facebook had to ditch the V8 JIT for their purpose it's totally understandable why Sciter chose QuickJS as their new Javascript engine.