NVidia sass disassembler/inline patcher

Pwning Santa before the bad guys do 🎅

A dynamic binary instrumentation tool for tracing and analyzing CUDA kernel instructions.

MoE training for Me and You and maybe other people

The MATLAB Tensor Core: a set of models of tensor cores written in MATLAB

Fast arithmetic modulo `2^k`, `2^k - 1`, and `2^k - d`.

Unofficial description of the CUDA assembly (SASS) instruction sets.

Low-overhead tracing of all Linux kernel-user transitions, for serious performance analysis. Includes kernel patches, loadable module, and post-processing software. Output is HTML/SVG per-CPU-core …

Awesome Object Capabilities and Capability Security

A UNIX-like kernel for the i386 architecture

A fast, small C/C++ function call tracer for x86-64/Linux, supports clang & gcc, ftrace, threads, exceptions & shared libraries

TexLive programs bundled into a single static binary for x86_64-linux / WASM

Inspect and dissect an ELF file with pretty formatting.

My solutions for CTF challenges

LLM inference in C/C++

High-efficiency floating-point neural network inference operators for mobile, server, and Web

FB (Facebook) + GEMM (General Matrix-Matrix Multiplication) - https://code.fb.com/ml-applications/fbgemm/

Transformer related optimization, including BERT, GPT

CUDA Templates and Python DSLs for High-Performance Linear Algebra

Open standard for machine learning interoperability

ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator

Tensors and Dynamic neural networks in Python with strong GPU acceleration

BLAS-like Library Instantiation Software Framework

oneAPI Deep Neural Network Library (oneDNN)

OpenBLAS is an optimized BLAS library based on GotoBLAS2 1.13 BSD version.