Nadav Rotem

bistra. Bistra is a domain-specific language designed to generate high-performance kernels (such as GEMMs, convolutions, etc). The program is designed to allow powerful compiler optimizations and code generation that are not possible in C. The tool can auto-tune GEMM kernels to around 90% of peak performance (on X86/AVX2) within seconds.

tensorli. Absolute minimalistic implementation of a GPT-like transformer using only numpy (<650 lines).

compressor. An educational implementation of a modern compressor in Rust

raisin. A simple lightweight set of implementations and bindings for compression algorithms written in Go.

dynamic-markov-compression. Dynamic Markov Compression (DMC) implemented in Rust. DMC is a lossless, general purpose, compression technique based on a probabilistic model.

Dynamic-Marcov-Compression. DMC

wasm-dot-renderer. HTML

rjvm. A tiny JVM written in Rust. Learning project

pgo_ml. Source code for the paper "Profile Guided Optimization without Profiles: A Machine Learning Approach"

memset_benchmark. This repository contains high-performance implementations of memset and memcpy in assembly.

layout. Layout is a rust library and a tool that renders Graphviz dot files.

fast_log. A fast implementation of log() and exp()

arpfloat. An arbitrary-precision floating-point library in Rust

FiniteStateEntropy. New generation entropy codecs : Finite State Entropy and Huff0

glow. Compiler for Neural Network hardware accelerators

hermes. A JavaScript engine optimized for running React Native.

llvm-project. The LLVM Project is a collection of modular and reusable compiler and toolchain technologies.

code-for-blog. Code samples from my blog