← Kernels & Compilers

Core

Triton Programming Model

Triton expresses GPU work as operations on blocks of values. You choose shapes, addresses and masks while the compiler maps the work onto hardware.

This concept is included with a free account. Sign in to continue reading.

free account · no card needed

RELATED CONCEPTS

CUDA Programming Model Kernel Fusion Memory Coalescing Tiled Matrix Multiplication torch.compile and CUDA Graphs FlashAttention Internals

LESSONS THAT TEACH THIS

Kernels and the MachineA tiled matrix multiply connects ownership, reuse and synchronization18 min→Kernels and the MachineFusion removes traffic that tiling cannot19 min→

PRACTICE THIS IN REAL QUESTIONS

CUDA, Triton & Kernel EngineeringAn AI-generated kernel passes random tests and claims a 3x speedup. What would you verify?→Napkin Math, Cost & CapacityA kernel does 4 TFLOP and moves 40 GB in one call. On an H100, is it memory-bound or compute-bound?→CUDA, Triton & Kernel EngineeringWhen do you write a kernel in Triton, and when do you have to drop down to CUDA?→CUDA, Triton & Kernel EngineeringWrite a fused row softmax in Triton, explain its memory traffic, and say where it stops scaling.→Open-Weights Models & Serving EnginesWhich GPUs can run an FP8 or AWQ checkpoint, and what do you actually get on an older card?→Open-Weights Models & Serving EnginesYou upgraded the serving engine and throughput fell 15 percent. Find the cause.→