I am interested in large numbers of requests per second for highly performant systems.
I wrote a parallel imaginary assembly interpreter which can send integers and code jumps between threads. This is backed by an actor system that I wrote and integrated with the interpreter that can communicate between threads. The actor system can handle between 19-100 million requests per second depending on variation. The interpreter can handle 674,552 requests per second with method sending and 1703141 without. Some variations generate messages upfront, others in parallel and others as the program works.
I am working on a multithreaded language. I want to avoid data races similar to Rust.
If you are interested in parallelism and programming language design then send me an email. Maybe we can join Multiprocess community on Discord ran by eatonphil. That server has advanced groups for operating system development, programming languages and database internals.
Things I am thinking of lately:
- automatic parallelization of code
- python GIL and threading. I need a method of sending complicated objects between interpreters that doesn't involve marshalling. I think I need to write an custom block allocator.
I also journal my computer software ideas in the open on GitHub