6 weeks into performance tuning and chasm is reall...

# webassembly

[Charlie Tapping]

6 weeks into performance tuning and chasm is really starting to shape up, the runtime is orders of magnitude faster than it was back in January and if you’re building wasm binaries with GC instructions you can expect to see over 100x speedup. I’m now able to run large complex programs in milliseconds, in some cases surpassing wasm interpreters written in low level languages. That being said we’re not done with performance ahead the 1.0, the next phase of work I expect to have 4-5 x improvement on chasms dispatch loop. I’ve learned so much in the last month alone, writing an interpreter in a high level language with several compilers involved (thanks to KMP) presents a whole host of challenges that would simply not exist if I were working in a low level language. Plenty of blog posts to come around these learnings once all is said and done.

It’s also worth noting you can now run programs compiled to WASI P1 with chasm thanks to a set of host functions provided by @Alexey Illarionov in the library here.

Road to 1.0 is roughly

• Performance: Bytecode fusion and final pass over instructions
• Functionality: Threads Proposal
• API: Functions to deal with Strings
• Usability: A gradle plugin to generate a pure kotlin interface for chasm, hiding the low level api for the majority of consumers
Latest version here