Skip to main content
Ctrl+K
NVIDIA CUTLASS Documentation - Home NVIDIA CUTLASS Documentation - Home

NVIDIA CUTLASS Documentation

NVIDIA CUTLASS Documentation - Home NVIDIA CUTLASS Documentation - Home

NVIDIA CUTLASS Documentation

Table of Contents

  • Changelog

CUTLASS Python

  • Overview
  • Quick Start Guide
  • DSL Programming Model
    • Introduction
    • Code Generation
    • Control Flow
    • JIT Argument Generation
    • JIT Argument: Layouts
    • Struct-like JIT Arguments
    • JIT Caching
    • JIT Compilation Options
  • Guides
    • Educational Notebooks
    • MMA Programming Guides
      • Warp-Level MMA Instructions Programming Guide
      • Warpgroup MMA Programming Guide
      • tcgen05 MMA Programming Guide
    • Integration with Frameworks
    • Debugging
    • In-Kernel Event Tracing (IKET) Profiling
    • Autotuning with the DSL
    • Ahead-of-Time (AOT) Compilation
    • TVM FFI Compilation
  • API Reference
    • Basic Data Types
    • Primitives (experimental)
    • CUDA (Jittable)
    • CuTe
      • Core
      • Runtime
      • Math
      • GPU Operations
        • nvgpu
        • nvgpu.common
        • nvgpu.warp
        • nvgpu.warpgroup
        • nvgpu.cpasync
        • nvgpu.tcgen05
      • arch
      • Naming Conventions
    • Pipeline
    • Task Scheduling (experimental)
      • Introduction
      • Resources
      • Pipeline Types
      • Schedules
      • Scheduling Patterns
      • Validation
      • Reading the Printed Output
      • Allocators
      • Pipeline Groups
      • Programmatic Dependent Launch
      • API Reference
        • task_scheduling
        • resources
        • task
        • task_manager
        • schedule_builder
        • memory
        • pipeline_group
        • pipeline
        • enums
    • Utilities
      • Hopper (SM90)
      • Blackwell (SM100)
  • Talks and Presentations
  • Limitations
  • Deprecation Policy
  • FAQs

CUTLASS Operator API

  • CUTLASS Operator API
  • Tutorials
    • Core concepts and basic GEMM
    • GEMM with fused epilogue
    • Bringing your own kernel
    • Host latency best practices
    • Fake tensors for compilation
    • Grouped GEMM with contiguous offset
    • Block-scaled GEMM (MXFP8)
  • API reference
    • Operator interface
    • Arguments and Operands
    • Kernel discovery
    • Metadata
    • Miscellaneous

CUTLASS C++

  • Overview
  • Getting Started
    • Quickstart
    • IDE Setup
    • Build
      • Building on Windows with Visual Studio
      • Building with Clang as host compiler
    • Functionality
    • Terminology
    • Fundamental Types
    • Programming Guidelines
    • GEMM Heuristics
  • Efficient GEMM in CUDA
  • Synchronization primitives
  • CUTLASS Profiler
  • GEMM Performance Measurement Guidelines
  • Dependent Kernel Launch
  • Blackwell Specific
    • Blackwell SM100 GEMMs
    • Blackwell Cluster Launch Control
  • CuTe
    • 00_quickstart
    • 01_layout
    • 02_layout_algebra
    • 03_tensor
    • 04_algorithms
    • 0t_mma_atom
    • 0x_gemm_tutorial
    • 0y_predication
    • 0z_tma_tensors
  • CUTLASS 3.x
    • Design
    • GEMM Backwards Compatibility
    • GEMM API
  • CUTLASS 2.x
    • Layouts and Tensors
    • GEMM API
    • Tile Iterator Concepts
    • Utilities
  • Code Organization
  • Grouped Kernel Schedulers
  • CUTLASS Convolution

Reference

  • Software License Agreement
  • Guides

Guides#

  • Educational Notebooks
  • MMA Programming Guides
  • Integration with Frameworks
  • Debugging
  • In-Kernel Event Tracing (IKET) Profiling
  • Autotuning with the DSL
  • Ahead-of-Time (AOT) Compilation
  • TVM FFI Compilation

previous

JIT Compilation Options

next

Educational Notebooks
NVIDIA NVIDIA

Copyright © 2025-2026, NVIDIA Corporation.

Last updated on Aug 06, 2026.