# Triton Crash Course: Block Pointer + Autotune + Masks — 50 Satırda GPU Kernel

> Source: https://sukruyusufkaya.com/learn/fine-tuning-cookbook/ftc-triton-crash-course
> Updated: 2026-08-21T21:46:29.792Z
> Category: Fine-Tuning Cookbook (Model-by-Model)
> Module: Part XIII — Custom Kernels & Performance Surgery
**TLDR:** Triton (OpenAI, 2021) — CUDA kadar hızlı, Python kadar kolay GPU kernel framework'ü. \`@triton.jit\`, \`tl.program_id\`, \`tl.arange\`, block pointer arithmetic, autotune decorator, mask-based load/store, shared memory abstraction. RTX 4090'da Triton vector add → matmul → softmax kernel'larını sıfırdan yaz.