3 ms·

I’m very fascinated by this, and I hope that your proposal gets approved! I’m a community college instructor in Silicon Valley, and my plan this summer (which

I’m very fascinated by this, and I hope that your proposal gets approved! I’m a community college instructor in Silicon Valley, and my plan this summer (which is when I have nearly three months off from teaching) is to work on a side project involving domain-specific languages for systems software. I’ve been inspired by the STEPS project, and I dream of systems being built with higher levels of abstraction, with very smart compilers optimizing them.

morphle 2y ago
Why not collaborate, it will help you avoid reinventing some wheels. For example make a 3D version of the 2.5D graphics Nile/Gezira. You could do it in less than 500 lines of code and within 3 months. Other system software could be a new filesystem in 600 lines of code or a TCP/IP in 120 LOC. I also think a SPICE or physics simulator could be around 600 lines of code. I'll do the parallelizing optimizing adaptive compilers and autotuners (in Ometa 2). I intend to target a cluster of M3/M4 Macs with 32-core CPU, 80-core GPU and 32-core Neural Engine cores with an estimated 80 trillion TOPS and 800 Gbps memory bandwidth each. A smaller $599 base model M4 Mac mini would do between a fifth and a third of that performance. Together we could beat NVDIA's complexity and performance per dollar per Watt in a few thousand lines of code.