Results 14 comments of Evan Mays

After finding Vivado's implementation & timing buttons, I was wrong about executing in one cycle. Still, I was able to get the math working for a mathematically guaranteed 1 instruction...

I find it interesting that all vliw implementations require the compiler to optimize the memory hierarchy and function unit utilization. Optimizing memory accesses can be done locally. After hashing all...

With 5 functional computational units (unopALU, binopALU, reduceALU, matmul, mulacc), a newbie writing the VLIW instructions in python wouldn't be too much different from the existing cherry.py code in tinygrad....

This net adds LOC w/o adding new features and removes the ability for anyone to use our keys :( We'd like people to use the keys until the free credits...

An obscure meme. The origin of the format can be found by looking up "Apple commercial what's a computer" on YouTube ![whats-a-backend](https://user-images.githubusercontent.com/20430648/214383306-5c049651-496d-4c4a-bda4-85d44c869afe.jpg)

Good catch See https://github.com/TheAppleTucker/backend-GPT/pull/2

Interesting, I'm on M2 and didn't go through those steps. Pretty sure I don't have those extra deps they ask for since my M2 is new Fwiw, Ray is removable....

PR #17 should remove `ray` dependency

@alexloi Ideally we can keep the idea of "database as a json file" for ease of use. python std lib should provide everything needed for locking and whatever else is...

No need for modal, that part of the readme should be deleted. Just the first 3 steps are enough. +dependency installation