Stories by skidrow

DeepSeek-R1 and FP8 Mixed-Precision Training

How to Write a Fast Matrix Multiplication from Scratch with Tensor Cores (2024)

DeepSeek-R1 and FP8 Mixed-Precision Training

Implementing a Fast Tensor Core Matmul on the Ada Architecture

How to Write a Fast Matrix Multiplication from Scratch with Tensor Cores

Understanding Peak, Max-Achievable and Delivered FLOPs

DeepSeek-R1 and FP8 Mixed-Precision Training

Outperforming cuBLAS on H100: A Worklog

Optimizing Matrix Multiplication on RDNA3

Outperforming cuBLAS on H100: A Worklog

Mastering LLM Techniques: Inference Optimization

Optimizing Matrix Multiplication on RDNA3

Outperforming cuBLAS on H100: A Worklog

Understanding Latency Hiding on GPUs [pdf]

AMD Radeon RX 9070 Series Linux GPU Compute Performance

Outperforming cuBLAS on H100: A Worklog

GPU Gems

Understanding Latency Hiding on GPUs [pdf]

A guide to LLM inference and performance

1