Skip to main content

Ctrl+K

Overview

Getting Started

Quickstart
CUTLASS Terminology
IDE Setup for CUTLASS Development
Functionality

Core Concepts

Fundamental Types
Layouts and Tensors
Tile Iterator Concepts
Synchronization primitives
CUTLASS Code Organization
Programming Guidelines
CUTLASS Utilities

Key Operations & APIs

CUTLASS GEMM API
CUTLASS 3.0 GEMM API
Efficient GEMM in CUDA
CUTLASS Convolution

CUTLASS 3.x Specifics

CUTLASS 3.0 Design
CUTLASS 3.0 GEMM Backwards Compatibility

CUTE - Compositional Universal Tile Engine

Getting Started With CuTe
CuTe Layouts
CuTe Layout Algebra
CuTe Tensors
CuTe Tensor algorithms
CuTe’s support for Matrix Multiply-Accumulate instructions
CuTe dense matrix-matrix multiply tutorial
Predication: What to do when tiling isn’t perfect
CuTe TMA Tensors

Advanced Features & Platform Specifics

Dependent kernel launches
CUTLASS Grouped Kernel Schedulers
Blackwell SM100 GEMMs
Blackwell Cluster Launch Control
CUTLASS Profiler

Building CUTLASS

Building on Windows with Visual Studio
Building with Clang as host compiler

Repository
Open issue

Index

By Nvidia