Skip to main content
Ctrl+K

tvm-tirx

Get Started

  • Quick Start

IR Guides

  • IR Overview
  • Types and Expressions
  • Statements
  • Operators, Calls, and Regions
  • Functions and Modules
  • Printing and Parsing
  • Structural Traversal, Rewriting, and Comparison
  • Symbolic Integer Analysis
  • Working with IR in C++
  • Working with IR in Rust

Programming Guides

  • Programming Overview
  • Language Basics
  • Control Flow
  • Types, Tensors, and Memory
  • Metaprogramming
  • Calling CUDA and PTX
  • Data Layout
  • Tile Operations
  • Compiling and Inspecting
  • Profiling with IKET

Extending TIRx: everything is a plugin

  • Getting Started with Rust Extensions
  • Defining Types, Operators, and Passes
  • Using and Compiling the Extension

Reference

  • Python API
    • tvm_tirx.runtime
    • tvm_tirx.ir
    • tvm_tirx.ir.prim
    • tvm_tirx.ir.transform
    • tvm_tirx.sym
    • tvm_tirx.script
    • tvm_tirx.script.parser
    • tvm_tirx.script.ir_builder
    • tvm_tirx.target
    • tvm_tirx.tirx
    • tvm_tirx.tirx.op
    • tvm_tirx.tirx.analysis
    • tvm_tirx.tirx.transform
    • tvm_tirx.tirx.layout
    • tvm_tirx.tirx.tile
    • tvm_tirx.tirx.script
    • tvm_tirx.codegen
    • tvm_tirx.backend
    • tvm_tirx.backend.cuda
      • tvm_tirx.backend.cuda.axis
      • tvm_tirx.backend.cuda.backend_config
      • tvm_tirx.backend.cuda.codegen
      • tvm_tirx.backend.cuda.compiler
      • tvm_tirx.backend.cuda.iket
      • tvm_tirx.backend.cuda.launch
      • tvm_tirx.backend.cuda.ptx
      • tvm_tirx.backend.cuda.thread_scope
      • tvm_tirx.backend.cuda.tile
      • tvm_tirx.backend.cuda.BackendConfig
      • tvm_tirx.backend.cuda.export_cuda_host
      • tvm_tirx.backend.cuda.register_backend
      • tvm_tirx.backend.cuda.script_namespaces
      • tvm_tirx.backend.cuda.script_namespace
    • tvm_tirx.backend.cuda.op
    • tvm_tirx.backend.cuda.script
    • tvm_tirx.backend.cuda.layout
    • tvm_tirx.backend.cuda.lang
    • tvm_tirx.backend.cuda.transform
    • tvm_tirx.backend.rocm
    • tvm_tirx.backend.metal
    • tvm_tirx.backend.metal.op
    • tvm_tirx.backend.opencl
    • tvm_tirx.backend.adreno
    • tvm_tirx.backend.hexagon
    • tvm_tirx.backend.vulkan
    • tvm_tirx.backend.webgpu
    • tvm_tirx.backend.trn
    • tvm_tirx.backend.trn.op
    • tvm_tirx.backend.trn.transform
    • tvm_tirx.backend.rust
    • tvm_tirx.support
    • tvm_tirx.testing
    • tvm_ffi structural utilities
    • tvm_tirx.compile
  • C++ API
  • Rust API

Developer Manual

  • Build from Source
  • Build This Doc Site
  • CUDA Backend Development
  • Reproduce CI/CD
  • Release Process
  • .rst

tvm_tirx.backend.cuda

Contents

  • Namespaces
  • Classes
  • Functions

tvm_tirx.backend.cuda#

CUDA-owned TIRx modules.

Namespaces#

  • tvm_tirx.backend.cuda.op

  • tvm_tirx.backend.cuda.lang

  • tvm_tirx.backend.cuda.layout

  • tvm_tirx.backend.cuda.script

  • tvm_tirx.backend.cuda.transform

  • tvm_tirx.backend.cuda.axis

  • tvm_tirx.backend.cuda.backend_config

  • tvm_tirx.backend.cuda.codegen

  • tvm_tirx.backend.cuda.compiler

  • tvm_tirx.backend.cuda.iket

  • tvm_tirx.backend.cuda.launch

  • tvm_tirx.backend.cuda.ptx

  • tvm_tirx.backend.cuda.thread_scope

  • tvm_tirx.backend.cuda.tile

Classes#

BackendConfig

Optional CUDA overrides.

Functions#

export_cuda_host(mod)

Return C++ host source with embedded device binaries.

register_backend()

Register CUDA-owned Python semantics.

script_namespaces(**_)

Return CUDA-owned TVMScript namespaces.

script_namespace(**kwargs)

Return the CUDA TVMScript namespace object.

previous

tvm_tirx.backend.load

next

tvm_tirx.backend.cuda.axis

Contents
  • Namespaces
  • Classes
  • Functions

By Apache TVM Community