Appendix · Further reading
Start here
Appendix

Further reading

The papers, guides, and repositories worth your time

The primary sources. Anything worth understanding properly is better read there than summarized here, and most of these are readable in an evening.

Start here

The minimum mental model, read in order before anything else.

Programming model

Threads, warps, memory hierarchy, and the CUDA execution model.

Compilation and machine code

What CUDA compiles down to, and how to read it.

Foundational kernel exercises

The exercises everyone works through first: transpose, reduction, scan, softmax.

Matrix multiplication

Tiling, layouts, and the reference GEMM implementations.

Direct implementation work

Worklogs that build a fast kernel from scratch, in the open.

Tensor cores and low precision

FP8, FP4, and the formats and instructions tensor cores run.

Attention

FlashAttention and the kernels it led to.

Triton

The blocked-program language and compiler.

CUTLASS, CuTe, and CUDA Tile

NVIDIA's tile-based kernel-authoring stack.

Other hardware stacks

Kernel programming models for AMD, TPU, and Trainium.

Profiling, benchmarking, and correctness

Tools for measuring and verifying what a kernel actually does.

Scheduling and continuous batching

Iteration-level scheduling and the engines that implement it.

KV cache systems

Shrinking, quantizing, and distributing the KV cache.

Quantization

Weight and activation quantization for inference.

Speculative decoding

Drafting and verifying multiple tokens per step.

Structured decoding and fairness

Constrained generation and scheduling fairness.

Long context and multimodal inference

Sparse attention and multimodal engine support at long context.

Parallelism, collectives, and topology

Tensor and pipeline parallelism, and the interconnects underneath.

Mixture-of-experts serving

Routing, dispatch, and load balancing for MoE inference.

Prefill and decode disaggregation

Splitting prefill and decode across workers.

Production systems

Scheduling and routing for serving at production scale.

Serving benchmarks

Workloads and metrics for measuring serving systems.

NVIDIA

Current NVIDIA data center GPU architecture.

AMD

Current AMD data center GPU architecture.

Google TPU

Current TPU architecture and kernel programming model.

AWS Trainium

Current Trainium architecture and kernel programming model.

AI-generated kernels

Benchmarks for LLM-generated GPU kernels.

Watchlist

Hardware and techniques worth tracking before they have reproducible evidence.

123 sources

Links go to the primary sources; anything worth understanding properly is better read there than summarized here.